|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
ARE THERE CONTINGENT, A PRIORI TRUTHS?
DANIEL CARTER MCCAIN
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS
UNIVERSITY OF FLORIDA
Daniel Carter McCain
I would like to thank my chair, Dr. Greg Ray, for the myriad observations he made on
various drafts and for his tireless effort toward the completion of this project. I would also like
to thank my committee members, Dr. Kirk Ludwig and Dr. Michael Jubien, for their insightful
comments and helpful suggestions at my defense. In addition, without the support of my fiancee,
Karen Garito, this project would not have come to fruition. Finally, I could not imagine
completing a task of this difficulty without the help of my Lord Jesus Christ.
TABLE OF CONTENTS
A CK N O W LED G M EN T S ...................................................... ........................................ 3
A B S T R A C T ........................................... ............................. 6
INTRODUCTION ....................... ........ .. .. .... ..... ...................
K R IPK E A N C A N D ID A TE S ............................................................................. ......................10
B ack g rou n d ............................ ........... ................................................................10
The "Standard M eter B ar" E xam ple ......................................................................... ... ... 10
Problem s w ith the K ripkean Exam ples............................................................ ... ............ 11
C o n clu sio n .........................................................................19
A C TU A L IZED CA N D ID A TE S.......................................................................... ....................20
B ack g rou n d ......................................................... ... ....... .......................... ............... 2 0
Bostock's Argument Against the "Actual Inventor of the Zip" Example ............................22
'A ll A actual G eniuses are G eniuses'. ........................................................... .....................27
C o n clu sio n .........................................................................2 9
PAST OBJECTIONS CONSIDERED IN LIGHT OF
'ALL ACTUAL GENIUSES ARE GENIUSES' ................. ............ ....................... 30
What Does it Mean for a Candidate to be "Philosophically Interesting"? ...........................30
D onnellan's O objections R evisited......... ............................................... ..... ............... 31
Kitcher's Objection Revisited and Concluding Remarks...............................................34
GETTING CLEAR ON 'ALL ACTUAL GENISUES ARE GENISUES' ...............................35
C clarification and B background ......... .. ............. ............................................................35
A analyzing 'A ctual'..................3........ ........................................... 39
The Contingent A Priori and Knowledge of One's Actuality ................ .......... ..........42
Two-Dimensional Logic and the Contingent A Priori ................................................44
C on clu sion .................................................................................4 9
A NON-INDEXICAL CANDIDATE FOR THE
C O N TIN G EN T A PR IO R I ........................................................................... ......................... 50
Introduction ............. .............. .. ......... .................................50
Williamson and Oppy's Conversation ............. .... ......... .......... ............... 50
C o n clu sio n .........................................................................5 6
C O N C L U SIO N ............................................................................................57
L IST O F R E F E R E N C E S ............................................................................... ...........................60
B IO G R A PH IC A L SK E T C H ............................................................................... .....................6 1
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Arts
ARE THERE CONTINGENT, A PRIORI TRUTHS?
Daniel Carter McCain
Chair: Greg Ray
Major Department: Philosophy
While most philosophers have thought that necessity and priority are co-extensive,
there have been some who have challenged this claim. I examine the claims of Kripke, Fitch,
Bostock and Williamson to see whether the candidates they offer for the contingent apriori are
viable. I conclude that, for various reasons, each of them is not a true example of the contingent
apriori. However, I offer a candidate that avoids the pitfalls of these alleged examples: 'All
actual geniuses are geniuses'. I conclude the paper by exploring and clarifying the semantic role
of 'actual', in the process defending 'All actual geniuses are geniuses' as a contingent, a priori
From Immanuel Kant to philosophers of the present day, the possibility of a divergence
between necessity and priority has been denied. If Kant and those who agree with him are
right, there can be no contingent, apriori propositions. Though the weight of tradition and the
majority opinion of contemporary philosophers fall in line with Kant, there are dissenters. While
the case for the necessary aposteriori and that for the contingent apriori are related,
establishing one does not automatically establish the other. Therefore, I will focus on the case
for the contingent a priori.
In this paper I will survey the literature on the contingent apriori in order to accomplish
two things: 1) trace the course the conversation has taken as different candidates for the
contingent apriori have been refuted and new ones have developed, and 2) question whether it is
plausible that a viable candidate will ever be offered. We will consider a few different types of
The first sort of candidate considered will be the Kripkean examples. These depend on a
linguistic stipulation that specifies the referent of a name in all possible worlds. It is argued by
Kripke and others that proper names have the same referent in all possible worlds. Thus, if one
stipulates that the referent of a certain name is the individual or thing which satisfies a certain
definite description, while it will be true apriori that the definite description is true of and only
of the referent of the name, it will be a contingent truth since the definite description might have
been satisfied by someone else.
We will next consider the sort of example that Fitch labels the incorrigibilityy" type.
There are, allegedly, contingent truths that one cannot believe falsely. These examples will
depend on indexicality. For instance, 'I have a headache' and 'I exist' would arguably be in this
group. It is then argued that if one cannot believe these truths falsely, then if one knows them to
be true at all one knows them apriori.
We will next consider the sort of example that I'll label the "actualized" type. These
examples, like the Kripkean ones, depend on rigid designation to allow us to say things we can
know apriori that are nonetheless false in some possible worlds. The difference is, while
Kripkean proper names lack sense, actualized terms do not. The "rigidification" is accomplished
not by linguistic stipulation, but by the use of 'actual' and its variants. For instance, 'actual
receptionists' has a discernable sense, but rigidly designates all and only those things which are
Finally, we will examine a non-indexical candidate for the contingent apriori. Examples
of this sort could also be labeled the "self-satisfying" type. These examples are argued to be
made true simply by believing them to be true. The trick is that they will be justified not by a
deduction utilizing indexicality (one's knowledge of one's existence or one's belief states), but
by an allegedly indexical-free, belief-forming mechanism.
I will begin with the discussion found in Kripke's Naming and Necessity, which broke
from traditional thought on the subject and offered the first candidate for the contingent apriori.
I will then examine Keith Donnellan's response to Kripke in Donnellan's paper "The Contingent
A Priori And Rigid Designators". In "A Priority and Necessity" Philip Kitcher argues that part
of Donnellan's analysis is ill-formed, but agrees with his conclusion that there are no
philosophically interesting versions of the contingent apriori. I will also examine G.W. Fitch's
1 There are two things that need to be said here. First, I am not claiming that the meaning of 'actual' is unambiguous.
Rather, I will treat 'actual' in the same way it is treated in the papers I discuss (which will become clear later), so as
to shed light on the conversation and see what merit it has. Second, I believe 'actual' is a rigidifier in the following
sense: the phrase composed of 'actual' and the term it modifies refers to the same individuals) no matter where the
phrase occurs in a sentence (e.g. 'Actual receptionists' refers to the same individuals in 'All actual receptionists have
desks' as it does in 'Necessarily, all actual receptionists have desks').
argument against Kripke and his dismissal of a couple of other alleged contingent, apriori truths
in "Are There Contingent A Priori Truths?"
As these papers present certain problems for the Kripkean sort of candidates for the
contingent apriori, I will next consider candidates involving 'actual' and Bostock's attempt to
refute them in his paper "Necessary Truth and A Priori Truth". Bostock's comments here will
lead me to suggest a different example from the one he considers, one that I believe meets his
objections. I will also reconsider Donnellan and Kitcher's objections in light of this new
candidate, as well as the argument given by Albert Casullo.
I will then investigate in what way the two-dimensional modal logic of Davies and
Humberstone comes to bear on the question of candidates of the contingent apriori that involve
'actual' and its variants. I will also examine Williamson's attempt to offer an indexical-free,
contingent, apriori truth and see whether Graham Oppy's attempted rebuttal is effective. I will
then conclude by recapping the status of the various candidates and summarizing what we have
learned about the contingent apriori.
To properly understand Kripke's alleged examples of contingent, apriori truths, we must
remind ourselves of Kripke's thesis that proper names are rigid designators. He writes, "Let's
call something a rigid designator if in every possible world it designates the same object, a
nonrigid or accidental designator if that is not the case."1 We already have the intuition, says
Kripke, that when we talk about how a thing might have been we talk about a possible situation
for that thing, rather than talking about a different thing that is very similar to the original. Thus,
our use of proper names should rigidly designate so that our talk of possible worlds captures our
intuitions about counterfactuals. It is necessary that nine is greater than seven because both
'nine' and seven' are rigid designators that pick out the same object in every possible world.
'The number of planets', however, is a description rather than a name. Thus, the phrase
will designate whatever happens to be the number of planets in the world in question, not
necessarily the number of planets in the actual world. Traditionally, names have been considered
abbreviated definite descriptions, picking out whatever entity, if any, matches the description in
the world in question. Kripke believes, however, that definite descriptions are used to fix
reference of names, to make them rigid designators; names are not simply abbreviated definite
descriptions. If they were we would be asking about some different, albeit very similar,
individual from Alex when asking, "What if Alex had not been an engineer?"
The "Standard Meter Bar" Example
One of the interesting results Kripke believes this thesis provides (and one of the reasons
it has been so hotly contested) is that it seems to allow us knowledge of contingent, apriori
1 In order to avoid trivial counter-examples, Kripke only requires that the rigid designators pick out the same object
in the possible worlds in which the object exists.
truths.2 One of Kripke's examples of this goes as follows. Let us stipulate that 'one meter'
refers to the length of S, where S is a certain stick or bar in Paris. Now consider the proposition
'S is one meter long'. What are the epistemological and metaphysical statuses of this proposition?
Given that the definition of 'one meter' is "the length of S (at the actual world)," the proposition
seems to be straightforwardly apriori. However, it is surely a contingent fact that S is the length
it in fact is; we can easily imagine possible worlds in which S is longer or shorter than it actually
is. Therefore, the proposition is only contingently true, though still apriori.
An important thing to examine is how the sentence is being evaluated at merely possible
worlds. Do we take what the sentence expresses at the actual world and evaluate it at different
possible worlds, or do we evaluate the sentence according to what it expresses in each possible
world? Kripke believes we must do the former because since names are defined not by synonym
or abbreviated description but by fixing the reference, the reference of the names at the actual
world is what underlies our investigations of modality. Thus, the proposition is contingent, but
will be apriori only in the actual world. In every merely possible world the proposition would
assert that a stick in the actual world is the same length as a stick in the merely possible one,
something we cannot know apriori. If, however, we choose to evaluate the proposition in the
latter way, it will be necessary and apriori since in every possible world the proposition will be
equivalent to something like 'S (at this world) is the same length as S (at this world)'.
Problems with the Kripkean Examples
A worry that one might have with Kripke's example is revealed by the discussion of
modal analysis found above. There seems to be an odd sort of circularity between the definition
of 'one meter' and the proposition in question. 'One meter' is defined as "the length of S," but
the proposition asserts that S is one meter long (i.e., that S is the length of S). While it does not
2 See pages 54-57 of Naming and Necessity for the discussion summarized here.
seem problematic, the example seems somewhat contrived given that we do not normally assert
propositions of this form, and the circularity could mask some semantic problem. It is unclear
how strong this objection is, but for two reasons we shall not examine it further. First, there are
other, clearer objections we shall discuss that make this example untenable. Second, Kripke has
other examples of alleged contingent apriori truths that do not involve this same sort of
oddness.3 Therefore, even if an "argument from oddness" can be made against the "Standard
Meter Bar" example, it will not defeat Kripke's general assertion that his theory of proper names
opens the possibility for contingent, apriori truths.
In "The Contingent A Priori And Rigid Designators," Keith Donnellan responds to
Kripke's theory of names and to the possibility of contingent a priori truths.4 He formulates one
sort of "uneasiness" with Kripke's examples, writing,
"It might be put roughly as follows: If a truth is a contingent one then it is made true, so
to speak, by some actual state of affairs in the world that, at least in the sorts of examples
we are interested in, exists independently of our language and our linguistic
It is difficult to see how definition by stipulation, on which Kripke's examples depend, can give
us epistemic access to non-linguistic states of affairs. The position Donnellan defends within
this paper is that while we have no reason to deny that definition by stipulation is theoretically
viable, such a procedure does not produce "interesting" contingent apriori truths.
3 'Neptune is the planet causing such and such discrepancies in the orbits of certain other planets' is one example of
an allegedly contingent a priori truth that does not involve such oddness. This example can be found on pg. 79 of
Naming and Necessity. It is similar to the "Standard Meter Bar' example in that it relies on stipulation.
4 Contemporary Perspectives in the Philosophy of Language, eds. French, Uehling, and Wettstein. University of
Minnesota Press, 1980 (pgs. 45-60)
5 ibid. (pg. 46)
Donnellan begins by defending Kripke from Dummett's attack on the idea that names can
be introduced as rigid designators.6 He does so by making a distinction between arguing for
certain historic examples of names being introduced as rigid designators and arguing that it is in
principle possible to introduce names in this way. He argues that it would be impossible, without
an explicit stipulation, to tell if historic cases (such as Leverrier's 'Neptune') are actually
examples of rigid designation. Donnellan believes, however, that the philosophical worries
about the contingent apriori are equally strong even if introducing names as rigid designators is
merely possible. As he sees no reason to deny the possibility, he next shows how such a
procedure does not yield philosophically interesting, contingent, apriori truths.7
Donnellan argues that any examples of contingent apriori truths, if they are examples at
all, that follow from the stipulative introduction of a name as a rigid designator will not be
worrisome (unlike Kripke's alleged examples).8 He begins by noting the distinction between
knowing that a sentence expresses a truth and knowing the truth it expresses. For instance, I
know that the sentence 'Fhqwhgads are Fhqwhgads' expresses a truth since a thing must be itself.
If I do not know what Fhqwhgads are, however, I do not know the truth the sentence expresses.
He then writes that we will represent the (Kripkean) stipulative introduction of a name with:
(a) Provided that the $ exists, let "N is the D" express a contingent truth,
where 'the V' is a definite description and 'N' is a name.
Donnellan believes that using this locution helps bring certain relevant features to light
(which we will see later). He also argues that while it may seem worrisome that we are
stipulating that something is contingently true, it is not suspect. So long as we pick a name that
6 ibid. (pgs. 47-50)
7 The idea that any examples of contingent a priori truths will be philosophically uninteresting is shared by many of
those who disagree with Kripke (including Donnellan, Kitcher, and Bostock). I will explore later what
'philosophically interesting' may mean and whether the claim made by Donnellan and others is true.
8 ibid. (pgs. 51-8)
is not already in use within the language, we are not attempting to make some non-linguistic
state of affairs the case; rather we are simply creating a purely linguistic state of affairs.
Furthermore, Donnellan asserts that we do not come to have knowledge of any (non-linguistic)
state of affairs by our act of stipulation.
Donnellan has us consider Kripke's "Neptune" example to illustrate his point.9 If the
Neptunians knew of Leverrier's stipulation, would they be justified in saying he had discovered
their planet as the cause of the perturbations? Could they, in their language and using their name
for their planet, say he knew their planet was the cause of the perturbations? Donnellan takes the
answers to these questions to be "No" and concludes that there is no knowledge of a non-
linguistic state of affairs given by the stipulation.
Donnellan also turns our attention to the classic "Newman-1" example, which he takes to
be analogous to the above example.10 He says that any knowledge resulting from the stipulation
must be de re, saying,
". .. it would have to be knowledge about an individual in the sense that there is (or will
be) an individual about whom we now know something and if that individual turns out to
be John we now know something about John."11
He says this is not simply because the propositions involve rigid designators; one could have de
dicto knowledge as a result of rigid designation as well. What is required is that the rigid
designators lack descriptive content in the propositions in question. It is his contention that this
sort of stipulation fails to produce knowledge de re.
9 See footnote 5 for the gist of the example.
10 'Newman-1 is the first child born in the 21st century,' where 'Newman-1' is stipulated to refer to the first child
born in the 21st century.
1 ibid. (pg. 54)
Having thus laid the foundation for his argument, Donnellan gives a rough
characterization of two principles that hold for a wide number of cases but fail for both the
"Neptune" and "Newman-l" examples. One of these principles is as follows:
"If an object is called by one name, say 'N,' by one group of people and by another name
by a second group of people, say 'M,' and if, in the language of the first group 'N is O'
expresses a bit of knowledge of theirs and if 'is u' is a translation of 'is O' into the
language of the second group then if the relevant facts are known to the second group,
they can say truly that the first group 'knew that M is D."12
Obviously, the Neptunians cannot say that Leverrier 'knew that Enutpen is the planet causing
such and such discrepancies in the orbits of certain other planets.' Donnellan concludes that
unless some other explanation can be given, Kripke's examples fail to meet these criteria
because they fail to produce knowledge of non-linguistic matters (the second criterion is similar
to the first, applying to temporal cases like Newman-1 rather than Neptune).
So while Leverrier could have known (perhaps even apriori) that the proposition
expresses a 1i nilh, he could not have known apriori the truth the proposition expresses.
Donnellan notes that some might still wish to claim that these cases give examples of the
contingent apriori, that one has apriori knowledge of a linguistic fact that could have been
otherwise. Donnellan concedes that this could be true, but says that such examples are not
philosophically interesting or worrisome, that they could be produced with names being
stipulated rigid designators, and that traditional definitions will yield similar results.
One might object that Donnellan's use of the two conditions is suspect for one reason or
another. First, he readily admits that his conditions allow certain counter-examples, one of
12 ibid. (pg. 55)
which is Kripke's "Hesperus/Phosphorus" example.13 If other of Kripke's examples can be
shown to have the same features that allow general counter-examples to the conditions,
Donnellan's case would be much weaker. Second (and on a related note), Donnellan leaves open
the possibility for some other explanation of why the examples fail to meet the criteria. If such
an explanation could be given, we would have less reason to deny the propositions in question
could produce knowledge of non-linguistic matters.
Philip Kitcher argues that we need not agree with Donnellan that any knowledge
produced by the stipulation must be de re.14 Consider the proposition 'If Shorty exists then
Shorty is a spy,' where 'Shorty' refers to the shortest spy. The corresponding belief must be
either de dicto or de re according to Donnellan (this is the alleged dilemma that Kitcher intends
to deny). It cannot be de dicto because that would require that 'Shorty' have descriptive content,
but the description was used merely to fix the reference. Therefore, it must be de re. However,
if the belief were de re I should be able to say, upon meeting the shortest spy, "Ah, I knew (a
priori) that you were the shortest spy." I cannot do this, so my belief cannot be de re. Therefore,
I do not have a priori knowledge of the proposition expressed.
Kitcher argues that contrary to Donnellan's argument, our belief in this case is de dicto.
"The fact that 'the shortest spy' was used to fix the reference of 'Shorty' does show that
the name 'Shorty' does not have a particular descriptive content. However, to use that
description to fix the reference of 'Shorty', I must intend to use 'Shorty' as an abbreviation for a
closely related description: 'Shorty' must abbreviate 'the shortest actual spy'."15
13 See footnote 22 of Donnellan's paper.
14 Philip Kitcher, "A Priority and Necessity," fromAustralasian Journal oJ I 1,il.. -.1 l, 58 (1980), pgs. 89-101
15 ibid. (pg. 91)
Thus, my de dicto belief is 'If there is a shortest actual spy then the shortest actual spy is a spy.'
Despite disagreeing with Donnellan's analysis, however, Kitcher agrees that while we might
have knowledge of contingent apriori truths through stipulation, they should not cause any
In "Are There Contingent A Priori Truths," G.W. Fitch considers and dismisses several
alleged examples of the contingent apriori, culminating in his dismissal of Kripke's example.16
The candidates he considers prior to considering Kripke's he calls the incorrigibilityy" examples.
The first he considers is offered by David Benfield: 'I have a headache' uttered by someone else
at some specific time.17 This is certainly contingent, but it is clearly not apriori. The
proposition requires experiential evidence to be justifiably believed. Secondly, Fitch notes that
examples of this sort purport special access to apriori knowledge; if Frank has a headache, he
would be the only person in a position to know this apriori.18 Fitch argues that no human has
special access to apriori knowledge (indeed, this would seem to show that the individual has
some sort of individuating experiences he or she is drawing from), so examples of this sort
cannot be contingent, apriori truths.
Fitch next considers an example by Alvin Plantinga. Consider 'I know that 7+5=12'.
While the contingency of this is clear, it's a priority is much less so. One may be tempted to
think that knowing the above truth requires knowing 'I believe that 7+5=12', and that this is
certainly known aposteriori. Plantinga argues that we cannot demand that to be apriori a truth
must be known without the use of any experience, since certain experiences (namely those
required to attain the concepts) are needed even to know '7+5=12'.
16 The Journal of CriticalAnalysis vol. 6, no. 4, Jan./Apr. 1977
17 Fitch, 119
18 ibid. (pg. 120)
Fitch points out that the difference between the two is that in the former experience acts
as a justification for belief, while in the latter it does not.19 While experience is certainly
necessary to know anything (even if only to acquire the concepts involved), the hallmark of a
priori knowledge is it either requires no justification or is justified purely on the basis of other a
priori truths. Plantinga's example satisfies neither of these.
The third example considered by Fitch is 'I exist,' also given by Plantinga. Fitch replies
that this too is justified on the basis of experience, albeit not any particular experience.
Furthermore, this is yet another example that purports special access to apriori knowledge (only
I can know apriori that I exist). For these reasons, this candidate for the contingent apriori
must be dismissed as well.
Fitch now moves on to Kripke's example, 'Stick S is one meter long'.20 Fitch argues,
very similar to Donnellan, that while those who define 'meter' as 'the length of stick S' know
that 'Stick S is one meter long' expresses a truth, they do not know the truth it expresses.21
Because a priority must be maintained, the stipulators must not have had any experience of S.
Thus they are not acquainted with it, they do not know its length relative to other things, etc.
Therefore, it is not clear, says Fitch, that they know what they're saying when they utter 'Stick S
is one meter long'. He writes,
"If we make the distinction between sentences and propositions, then it seems clear that
the fixers only know something about the language they helped to form, not about the
world the language is used to describe."
19 ibid. (pg. 121)
20 ibid. (pg. 122)
21 ibid. (pg. 123)
Thus, he dismisses Kripke's alleged example of the contingent apriori. He closes his paper by
remarking that while necessity and a priority are different conceptually (one is metaphysical
while the other is epistemic), we have no reason as of yet to believe they are not co-extensive.
So, what can we conclude about the Kripkean candidates for the contingent apriori?
The worry about the "Standard Meter Bar" example having an odd sort of circularity between the
stipulation and the proposition in question doesn't rule it out as a candidate. However,
Donnellan's worry is more problematic. While the problems with Donnellan's argument that
Kitcher and I point out seem to be real problems for him, his conclusion that Kripke's examples
are not philosophically interesting, a conclusion shared by Kitcher seems right.22 Fitch's
comments reveal the essential point in Donnellan's paper without Donnellan's unnecessary and
problematic conditions and without insisting that the knowledge produced by Kripkean examples
must be de re. Fitch's paper reveals the essential problem with the Kripkean examples: linguistic
stipulation alone cannot give us knowledge of a particular and, thus, cannot generate apriori
knowledge of contingent states of affairs involving that particular. It also notes in passing the
failure of candidates of the incorrigibility type.
22 What this claim amounts to, and how it informs and directs the discussion on contingent, a priori truths, will be
David Bostock puts forth yet another objection to the idea that there are contingent a
priori truths in his paper "Necessary Truth and A Priori Truth."1 He writes,
"I believe that, in the end, Kripke's two claims [that there are contingent apriori truths
and that there are necessary aposteriori truths] must be admitted to be correct, but that
does not have quite the significance that one is apt to suppose."2
His conclusion is similar to that of Donnellan. Kripke gets the results he claims but that they are
ultimately not of much importance. However, Bostock reaches it for very different reasons.
While Donnellan focuses on what sort of knowledge is produced by alleged examples of the
contingent apriori, Bostock considers the question of rigid designation from Kripke's theory of
proper names and argues that rigid designation does not give us philosophically important
examples of the contingent apriori.
Bostock begins by laying out some of the characteristics of the quantified, modal logic
that underwrite Kripke's notions of rigidity and necessity. The key notion is that of possible
world, a counterfactual situation where every proposition has a determinate truth value. A
possible or contingent proposition is one that is true in some possible world; a necessary
proposition is one that is true in every possible world.
Based on Kripke's conception of names as rigid designators Bostock introduces another
rigidifiedd" part of speech. By specifying the extension of a predicate at every possible world
one introduces a rigid predicate.3 To give an example that will arise later in Bostock's argument,
we could stipulate that the extension of 'cordate' in all worlds is to be all and only those
1 Mind, New Series, Vol. 97, No. 387 (Jul., 1988), pg. 343-379
2 ibid. (pg. 344)
3 ibid. (pg. 351)
creatures that actually have hearts. In doing this Bostock separates the concern over the
contingent apriori (and the necessary aposteriori) from the viability of Kripke's conception of
naming. Rigidifying is a legitimate process apart from the question of whether names (or
predicates for that matter) are actually rigid and so the concerns over the purported material
difference between necessity and priority are valid ones. Note that Bostock once again takes
the same line as Donnellan in arguing that even if we do not in fact introduce terms as rigid
designators or expressions, the worries over the material difference persist.
Bostock says that we have a set of words in the English language 'actual' and its
variants that modify certain words or phrases and make them "Kripkean" (or rigid).4 The
function of these words is to anchor whatever they modify to the actual world even if we are
considering counterfactual situations. For example, consider the proposition 'All actual geniuses
are geniuses'. If I ask whether this proposition is necessary, I am asking whether it is true in
every possible world that every thing that is a genius in the actual world is a genius in that world.
Within quantified modal logic an actuality operator, 'A', is used to accomplish what is
accomplished by 'actual' and its variants in natural language.5 'A' modifies a proposition, O,
such that 'A D' is to be interpreted as true at any world iff 'D' is true at the actual world. 'A'
can also modify predicates; 'A Fx' means 'x is actually F'. As with other modal operators, when
combined with other scoped operators the 'A' operator may take both wide-scope and narrow-
4 ibid. (pg. 355)
5 ibid. (pg. 356)
Bostock's Argument Against the "Actual Inventor of the Zip" Example
With the foundations thus laid, Bostock gives his argument that the use of 'A', while
permitting a material difference between necessity and truth in every possible world, also gives
us results that do not match with our intuitive notions of necessity and contingency.
"Let 'p' be any plainly contingent proposition, that happens to be true as things are, but
could easily have been otherwise. Then 'Ap' will be true. But then it follows further that
'Ap' will be true in all possible worlds, for it will be true in all worlds that it is true in
this world thatp. If truth in all worlds suffices for necessity, we must then conclude that
it is a necessary truth that Ap. But this is surely absurd. We cannot really turn a
contingent proposition into a necessary one by adding such qualifications as 'actually' or
'in fact' or 'as things are', and the correct conclusion to draw is evidently that the
criterion of truth in all possible worlds is no longer an adequate criterion of necessity."6
In order to remove the alleged counter-intuitive results of our working notion of modality
(and thus of our modal system), Bostock suggests a different way of handling 'actual' and its
variants. In what follows we will examine his system and his reasons for promoting it. This
method will, according to Bostock, allow us to represent sentences containing 'actually' without
giving counter intuitive results. He writes,
"The general idea is to treat 'actually' not as a new modal operator but as a device for
indicating relative scope, so that the word no longer occurs explicitly in our formal
language, where scope is shown differently, by order and bracketing."7
6 ibid. (pg. 357-8)
7 ibid. (pg. 360)
I will briefly explain the changes Bostock suggests and the advantages of making this revision. I
will then question Bostock's motivation for the change and some of the assertions he makes
about the effects of the change.
Bostock works with 'It is possible that the person who did actually invent the zip should
not have done so'.8 The "old" formalization of this proposition would look something like the
0Ox (AVy(Fy y = x) -Fx), where 'F' is 'invented the zip'.
Bostock, however, suggests the following "new" formalization:
Ex (Vy (Fy y = x) & 0 -Fx).
In the second formalization, 'A' is no longer present. Rather, the expression modified by 'A' in
the first formalization is taken outside of the scope of the other modal quantifier and 'A' is
removed. Thus, the modality is now a de re modality with the formalization read as,
'Concerning the person who did invent the zip, it is possible that he should not have done so'.
Bostock believes that there are several reasons to accept this treatment of 'actually'.9
First, this method can handle all instances of 'actually'. Second, Bostock argues that it is more
versatile than the previous system. He says that there are examples of propositions that the
original treatment of 'actual' and its variants cannot formalize, such as 'It is possible that all the
actual dogs should have existed and some other dogs as well'. Finally, this method eliminates
what Bostock takes to be counter-intuitive results (e.g. "philosophically interesting" necessary a
posteriori truths). He writes, "The consequences of this [change in our modal system] for the
divergence between necessary truth and apriori truth are clear: there is now no divergence or
at least none that is due to the word 'actually'." Before moving on to the possible problem with
8 ibid. This proposition underlies the potential candidate '
9 ibid. (pg. 360-362)
Bostock's move, let us examine his claim that his treatment allows us to formalize propositions
that cannot be formalized by the traditional treatment.
Bostock uses the example mentioned above to illustrate this point. He says the point of
the sentence is to assert that there is a world in which all actual dogs exist and other (i.e. non-
actual) dogs exist. The problem, however, is that we cannot refer to the possible world in
question when we treat 'actual' as a quantifier itself, so we have no way of evaluating the
proposition asserted. The example can be represented with the following:
3wax(xsw & Dxw & Vy(Dyw* -- yew & yfx)),
where 'w' is a variable ranging over possible worlds, 'x' ranges over possible objects, 'w*'
denotes the actual world and 'Dxw' means 'x is a dog in w'. Bostock notes that we can use
'O3x(Dx &' in place of 'HwHx(xsw & Dxw &'. We may also replace 'Vy(Dyw* -*' with
' Vy(ADy -*'. This suggestion could be used to try to represent the proposition in our usual
system of modal logic. However, once we do we are unable to translate 'yew' since doing so
would require asserting that the dogs that actually exist also exist in the worldwe are discussing.
We cannot do this because we have dropped the '3w', which we would have to refer back to in
order to make the assertion.
Bostock's method, on the other hand, can treat this sentence and those like it with ease.
Without removing 'A' we formalize 'It is possible that all the actual dogs should have existed
and some other dogs as well' as:
[3O(Vx(xsp +-+ y (y = x)) & 3x (Dx & oVy(ADy -y ysp & y x)),
where '0' is a rigid-predicate variable that ranges over the possible world discussed. Therefore,
the above formalization can be roughly translated as 'There is some 0 such that there is a
possible world in which everything in 0 is in that world and everything in that world is in P.
Furthermore, in that world there is some x such that x is a dog and all of the actual dogs are in
the possible world and are not identical with x'. Then, utilizing the unique role '0' plays in the
language, Bostock can eliminate 'A' and translate the proposition with the following:
aa(Vx(xsa Dx) & 3PO(Vx(xsp Hy (y = x)) &
Ex (Dx & VDy(ysa -y yb & y x)))).
Bostock admits that this method is somewhat cumbersome, but says that it allows us to translate
these sorts of sentences without needing a further device. We can translate these sorts of
sentences while still utilizing 'A', but it requires a further device, like the one by Peacoke that
Bostock discusses briefly.10
So there is a trade-off between having simpler expressions on the one hand and not
having to introduce a further device into our formal language on the other. Which one we would
deem the simpler revision would largely be a matter of preference, so Bostock's "argument from
simplicity" does not tip the scales toward one or the other. We must, therefore, examine his
motivation for suggesting the revision in the first place (i.e. that the traditional treatment of
'actual' and its variants leads to counter-intuitive results).
Bostock's argument that we need a different way of handling 'actual' and its variants
seems plausible on the surface. Consider the following proposition: 'All philosophers are pale'.
This proposition is both contingent (it is true in some worlds and not in others) and aposteriori;
it poses no problems for the idea that necessity and priority are materially equivalent.
However, suppose the proposition above is actually true and consider the proposition 'It is
actually the case that all '. It seems to me that Bostock believes we have the intuition that the
modal status of this proposition should be the same as the first. Since in our current system of
10 ibid. (pg. 361)
modal logic the second proposition comes out necessary, Bostock believes there must be a
problem with some components) of our system.
However, do we really have the intuition that these propositions should have the same
modal status? We certainly do have the intuition (and rightly so) that it is not essential to
philosophers that they be pale. That is, we believe the proposition 'All philosophers are
necessarily pale' is false, and hence also that 'It is actually the case that all philosophers are
necessarily pale' is false. However, this is the narrow-scope reading of the necessity operator; it
is clear from Bostock's talk of contingent propositions and 'A' being a propositional modifier
that he has in mind de dicto necessity. So, what are our intuitions about 'Necessarily, it is
actually the case that all philosophers are pale'?
I must confess that I do not have clear, pre-theoretic intuitions on the status of de dicto
modal statements that include 'A'; my intuitions on de re modal statements including 'A' are
reasonably clear though. However, I think there are a few reasons to think that Bostock's
conclusions about our intuitions and what we should do with our formal notion of necessity are
either questionable or mistaken. First, since the traditional "truth in all possible worlds" notion
of necessity works in so many cases, it would seem irresponsible to throw it out unless intuitions
are reasonably clear and widely shared. Second, I think there is serious doubt that intuitions on
this matter are in fact clear and widely shared. The fact there is continued debate would seem to
be evidence of this. Third, it is at least possible that Bostock's intuitions about the modal status
of 'Ap' are, unbeknownst to him, intuitions about the de re necessity claims.
At the end of his paper, Bostock defends what he believes is a true example of the
contingent apriori (378). His example is 'If there is exactly one object that falls under the
predicate, then it is necessary that there is no more than one', where the predicate in question is a
rigidified predicate. Bostock notes that if the antecedent is false, the conditional turns out to be
contingent. However, given the nature of rigidified predicates, we are able to know the
proposition is true apriori. Therefore, we have a contingent, apriori truth based on a Kripkean
language, though of a different sort than the ones Kripke mentions.
Bostock's example seems to avoid the worry with the Kripkean examples (i.e. that we
cannot understand them apriori). While rigidified predicates, like Kripkean proper names, lack
sense, since the predicate is mentioned rather than used it does not seem to present a problem for
the example; we can understand the proposition. However, since we cannot know apriori which
things fall under the predicate, and since the predicate does not have a sense, the proposition
concerns only a linguistic state of affairs. As we shall see later when we discuss what it is for an
example to be philosophically interesting, Bostock's example fails to be so. So, while the
example seems to be an example of the contingent apriori, it is not of the sort that interests us.
'All Actual Geniuses are Geniuses'
While Bostock focuses on the actuality operator as a propositional modifier, one could
offer a candidate for the contingent apriori that uses 'A' as a predicate modifier. So, even if
Bostock is correct that the propositional modifier use of 'A' gives us counter-intuitive results
when considered in conjunction with the wide-scope reading of one of the other modal operators,
that may simply give us reason to limit where we place 'A' rather than reform our method for
assessing necessity. I will conclude my argument against Bostock's position by showing that the
narrow reading of the actuality operator (the one that gives us the contingent apriori) does not
lead to counter-intuitive results, contingent apriori aside.
Consider 'All geniuses are geniuses'. This proposition is true and apriori. Now
consider 'All actual geniuses are geniuses'. This also seems to be a priori, but what is its modal
status? Consider the individuals who are geniuses in the actual world. Is it true that they are
geniuses in all possible worlds? Certainly not, since Albert Einstein might have been a man of
average or below average intelligence, so 'All actual geniuses are geniuses' is a contingent
proposition. According to Bostock, the function of 'A' in this position is to take the term it
modifies out of the scope of the other modal operator, so 'All actual geniuses are geniuses' could
The above formula comes out false using our traditional system of modal logic, having the same
truth conditions implicit in the possible worlds explanation above. Therefore, in this case there
is no divergence between our intuitions and our system and, thus, no need for a Bostockean
Bostock briefly addresses the possibility of rigidified predicates (3 51).12 As rigidified
predicates are similar to 'actual geniuses' in our example, we should look at his remarks.
Though Bostock's treatment of 'A' as a sentential modifier only includes the wide-scope reading,
his treatment of 'A' as a predicate modifier allows him to handle narrow-scope uses of 'A' as
well. Bostock says one can introduce rigidified predicates by stipulating what the extension of
the predicate is to be in every possible world. In our example, the term 'actual geniuses' is taken
to refer to all and only those things which are geniuses in the actual world. Thus, in all merely
possible worlds, the extension of the predicate is determined by the things that happen to be
geniuses in the actual world and which also exist in the world in question.
11 Bostock makes implicit use of the principal that a proposition, p, is necessary just in case p' is true.
12 It is noteworthy that while Bostock brings up the possibility of rigidified predicates, he does not involve them in
his argument, nor does he say that they give counter-intuitive results.
Bostock explicates some of the features of this sort of rigidity with an example (351-2).
'Cordate' is to refer to all and only those creatures which have hearts in the actual world.13
'Renate' is to refer to all and only those creatures which have kidneys in the actual world. Now
we have two rigidified predicates. Suppose (as is the case) that it is a contingent, aposteriori
truth that all those things which have hearts also have kidneys. It is, therefore, a necessary, a
posteriori truth that all cordates are renates. It is a contingent, apriori truth that all cordates
have hearts since in some possible worlds they will not.
Bostock believes that these fail to produce examples of the contingent apriori as well
since he believes that his arguments we discussed earlier cover these cases as well. As I have
responded to his earlier arguments already, I shall not revisit them here.
I admit that I have not decisively refuted Bostock's argument. What I have done, at best,
is question whether he is warranted in assuming the things he does and advocating a revised
modal system. I have shown that it may be the case that Bostock's criticisms of purported
contingent, a priori truths involving wide-scope uses of 'A' are ill-founded, and that narrow-
scope uses of 'A' can be used to form truths, like 'All actual geniuses are geniuses', that do not
fall prey to Bostock's objections.
13 Bostock also says that if something has a heart in the actual world and does not exist in some possible world, it is
still a cordate in that world. This makes it such that something is necessarily a cordate if it is one at all.
PAST OBJECTIONS CONSIDERED IN LIGHT OF 'ALL ACTUAL GENIUSES ARE
What Does it Mean for a Candidate to be "Philosophically Interesting"?
While Bostock's worries do not apply to our current candidate, there is another possible problem
with candidates for the contingent apriori that depend on 'actual' and its variants. The Kripkean
examples were said to be "philosophically uninteresting". Can the same be said for our current
First, let us consider what a "philosophically interesting" contingent, apriori truth would
be. Donnellan, Kitcher and Bostock all admit that uninteresting examples of the contingent a
priori can be generated. However, none of them explains clearly what would make a case
philosophically interesting or uninteresting. In reference to 'Provided the D exists, "t is the Q"
expresses a contingent truth', Donnellan says,
". .. I am not sure whether in the circumstances what sentences of form (A) express are
both contingent and apriori. But if they are they are harmless varieties of the contingent
a priori, examples of which we could produce without recourse to stipulations
introducing rigid designators."1
What would these be and why are they uninteresting? Shortly after the above quote, Donnellan
says that one might argue that sentences of form (A) are both contingent and a priori since they
assert something true about language that might have been false. Donnellan's paper also utilizes
a distinction between linguistic states of affairs and non-linguistic ones, so perhaps the
uninteresting examples of the contingent a priori are simply statements of a priori knowledge of
a contingent, linguistic state of affairs.
1 Contemporary Perspectives in the Philosophy of Language (pg. 56)
If this is what Donnellan, Kitcher and Bostock have in mind it is clear why they think
such examples are uninteresting. For one, they are ubiquitous; most (if not all) of our linguistic
stipulations might have been otherwise. Second, the contingent, apriori truth wouldn't be 'S is
one meter long', for instance. Rather, it would be "One meter' refers to the length of S'. Or, to
put it in a locution similar to Donnellan's, 'Provided there is something which is one meter, "S is
one meter long" expresses a contingent truth'. I say allegedly since one might contest the idea
that our knowledge of the linguistic stipulations is apriori.
So, what would be a philosophically interesting contingent a priori truth? If what is said
above is correct, we need apriori knowledge of a non-linguistic state of affairs. As I have
shown, Bostock has not given us a decisive reason to reject 'All actual geniuses are geniuses'.
This example (and those like it) do not depend on the Kripkean theory of naming and at least
seem to make reference to a non-linguistic state of affairs. Therefore, 'All actual geniuses are
geniuses' would seem to be philosophically interesting.
Donnellan's Objections Revisited
It is notable that Kripke's examples seemed to be about non-linguistic states of affairs
until they were probed by Donnellan. Might the same thing happen to our current candidate? At
the very least, we have a reason to reevaluate Donnellan's claim rather than simply taking it to
have provided a refutation of all possible candidates for the contingent apriori. The examples
Donnellan considered in his paper were all based on rigidified proper names or other sorts of
linguistic stipulations. As a result, the terms did not have a sense that contributed to the truth
conditions in any direct way; the sole function of the sense was to fix the reference. Because of
this, Donnellan could argue that while we possibly know apriori that the sentence expresses a
truth, we cannot know the referent of the rigidified term a priori and thus cannot know what
truth the sentence expresses. Examples involving 'actual' and its variants do have sense that
may do more than simply fix the reference. Therefore, we must see whether they are different
from the sort of examples Donnellan considers in a way that will allow us to know apriori the
truth they express. To do this, I will step through the various parts of Donnellan's argument with
'All actual geniuses are geniuses' as the sentence in question.
One reason Donnellan rejects Kripke's examples is that they fail to meet two criteria
Donnellan set up to test whether they assert any non-linguistic truth. However, there is a
problem applying these criteria to our current candidate. The criteria are designed for examples
involving proper names. Furthermore, the criteria cannot be easily adapted to the candidate
we're considering since to do so would require treating 'actual genius' as if it didn't have a sense.
So, Donnellan's criteria, whatever force they may have against the Kripkean examples, simply
do not apply to 'All actual geniuses are geniuses'.
When Donnellan is concluding his argument he attempts to show where the line of
reasoning from rigid designation to contingent a priori truths went wrong. In doing so he gives
two possible things that can be known as a result of a stipulation:
(A)Provided the D exists, "t is the Q" expresses a contingent truth.2
(B) Provided the P exists, t is the Q.3
If we were to adapt these to our current candidate, they would be:
(A') Provided the P exists, "The actual P is a Q" expresses a contingent truth.
(B') Provided the P exists, the actual P is a Q.
2 It is important to note that Donnellan does not commit himself to form (A) knowledge being a priori. He merely
says that ifwe have a priori knowledge, it is form (A) rather than form (B). It is highly questionable whether we
even have form (A) knowledge a priori.
The flaw in the reasoning, according to Donnellan, is that while rigid designation can give us
form (A) knowledge, it cannot give us form (B) knowledge. (A) is knowledge of a purely
linguistic state of affairs while (B) is knowledge of a non-linguistic state of affairs. Do we have
form (B) knowledge if we know that all actual geniuses are geniuses?
It seems that we do have form (B) knowledge if we know that all actual geniuses are
geniuses, however. This is revealed if we consider why Kripke's examples give us form (A)
knowledge but not (B) The way in which we can have form (A) knowledge without form (B)
knowledge is if 't' does not have a sense that contributes to the truth conditions. This allows us
to know that a sentence expresses a (contingent) truth without knowing the truth it expresses by
simply stipulating that it expresses a truth.4 If't' lacks a relevant sense and we do not know its
referent apriori, we do not know the truth expressed. 'Is an actual genius' is not lacking in
sense, however, so we do not have this problem.
One might make the point, however, that to have form (B) knowledge one must know (a
priori) the referent of 't' in order to know that D is true of him (i.e. one must have de re
knowledge). It is clear that Donnellan thought this was the case.5 It is also clear that we do not
have a priori knowledge about any particular entity if we know all actual geniuses are geniuses;
we do not even know who the actual geniuses are. Does this show that we do not have form (B)
knowledge in this case?
Kitcher does not think so. As has already been shown with the example of 'Shorty is a
spy', one can have a bit of de dicto knowledge about a non-linguistic state of affairs. The
mistake Donnellan made, according to Kitcher, is requiring particular descriptive content rather
than simply descriptive content. While do not have any knowledge of descriptive content about
4 Though this is not the sort of stipulation Kripke had in mind, Donnellan cashes out Kripke's examples in these
terms. (pg. 52)
5 Pg. 58 and elsewhere
the referent of 't' (de re), we do have knowledge of some of the descriptive content of 't' (de
dicto). Therefore, we can have form (B) knowledge despite our lack of de re knowledge.
Kitcher's Objection Revisited and Concluding Remarks
An interesting question arises from this, however. Kitcher, despite disagreeing with
Donnellan's analysis, agrees with the conclusion that only uninteresting contingent a priori truths
can be generated. How can this be given that it seems we can have contingent apriori
knowledge of non-linguistic states of affairs rather than simply linguistic ones?
Kitcher says several things about Donnellan's conclusion and the possibility of
interesting contingent a priori truths. He says that anyone can come to a priori knowledge of
contingent truths, but doing so requires linguistic acts that are "at odds with the standard
functions of language."6 Kitcher believes that even examples that have a non-linguistic
component are uninteresting because they require odd linguistic stipulations in order to express.
Given that our current candidate does not seem to involve this sort of stipulation, we have
avoided Kitcher's worry.
It would seem, therefore, that none of the objections to the contingent a priori considered
up to this point have provided a refutation of our current candidate. Donnellan fails to account
for the fact that one can have form (B) knowledge without having knowledge de re, and while
Kripke's examples do not provide this sort of knowledge, 'All actual geniuses are geniuses' does.
6 Kitcher, 100
GETTING CLEAR ON 'ALL ACTUAL GENIUSES ARE GENIUSES
Clarification and Background
Casullo begins his paper, "Actuality and the A Priori," by summarizing and addressing
an argument by Kitcher which, though not appearing explicitly in the paper we discussed earlier,
underlies and supplements the argument discussed earlier. Casullo begins by giving the
following condition, which Kitcher believes is a condition on apriori warrants:
a is an a priori warrant for X's belief thatp if and only if a is a process such that, given
any total sequence of experiences which would have enabled X to form the belief that,
(a) some process of the same type as a could produce in X a belief thatp, (b) if a process
of the same type as a were to produce in X a belief that p then that process would warrant
X's belief that p, (c) if a process of the same type as a were to produce in X a belief that p
Casullo asks what knowledge, if any, this condition gives us about our own actuality. To
see whether the knowledge of our own actuality is apriori, Casullo says we must ask questions
about what our knowledge would have been had our experiences been different. This fact, in
combination with Kitcher's view that 'actual' is an indexical, necessitates we determine whether
it is possible to believe in world w the truth expressed by 'I'm actual' in w*. If it is, then belief
in one's actuality is not apriori.
Kitcher argues that since this result would be extremely odd, we must deny that 'I'm
actual' expresses different beliefs at different worlds. Kitcher does not give up his belief that
'actual' is an indexical, but says that for the purpose of determining and applying a criterion of a
priority, the difference between the tokens of 'I'm actual' doesn't matter. Casullo notes that the
upshot of this view is that 'I'm actual' is true in any world in which I believe it, so condition (c)
is satisfied, and thus the knowledge of one's actuality appears to be apriori (392). He argues,
however, that Kitcher's last move is ad hoc.
Let's adopt the following notation. 'S' refers to the speaker. '[P](a)' means 'the
proposition expressed by sentence P in world a. 'p,[P](a)' means 'world 1 makes true the
proposition expressed by P in a. Finally, 'a' designates the actual world and 'w' some other
Casullo begins with the following summation of Kitcher's claims:
(1) The belief that S expresses in a by tokening 'I'm actual' = the belief that S expresses in w by
tokening 'I'm actual'
(2) ['I'm actual'](a) ['I'm actual'](w).
These claims are tantamount to saying that while the different tokens of 'I'm actual' express
different propositions, they express the same belief. Casullo sums up Kitcher's defense of this
(3) If the belief that S expresses in a by tokening 'I'm actual' the belief that S expresses in w
by tokening 'I'm actual', then the belief which S expresses at a by tokening 'I'm actual' is
false at w.
Casullo supplements (3) with:
(4) It is not the case that the belief which S expresses at a by tokening 'I'm actual' is false at w.
He agrees with Kitcher that (3) and (4) entail (1), but Casullo thinks (3) is problematic. Since all
that is needed to justify (4) is the trivial fact that one cannot be wrong about one's actuality,
Casullo grants (4) and focuses on the previous premise (393). To disprove (3), Casullo has us
grant (contrary to Kitcher):
(6) The belief that I express in a by tokening 'I'm actual' = ['I'm actual'](a)
(7) The belief that I express in w by tokening 'I'm actual' = ['I'm actual'](w).
Conjoin these and we get:
(8) 'The belief that I express in a by tokening 'I'm actual' the belief I express in w by tokening
Notice that (8) satisfies the antecedent of (3), so it would seem reasonable to conclude that S's
belief that he exists in a is false at w. This is an absurd conclusion, Casullo points out, since the
belief is true both at a and w. He briefly considers the counter-argument that individuals are
world-bound, but dismisses it because it would only be found persuasive by those who accept
Casullo now argues that Kitcher's account contradicts the very things he hopes to derive
from it, namely that knowledge of one's actuality explains how we can have knowledge of
contingent, apriori truths of the kind proposed by Kripke. Casullo has us consider the Shorty
examples mentioned earlier in this paper (If Shorty exists, then Shorty is a spy). Kitcher argues
that to use 'the shortest spy' to fix the referent of 'Shorty', 'Shorty' must be acting as an
abbreviation of 'the shortest actual spy' (394). Therefore, 'Shorty is a spy' is best represented by
something like the following:
(12) If there is a shortest actual spy, then the shortest actual spy is a spy.
To determine whether (12) is known a priori on Kitcher's account, we must examine the
proposition the sentence expresses in a and in w. Casullo has us grant that Shorty is a tall
basketball player in w, while Kareem is the shortest spy in w. Now, 'If Shorty exists, Shorty is a
spy' is clearly apriori. Kripke understands it as a de re belief, picking out Shorty in both a and
w. Thus, it is a contingent truth; at w, Shorty is a tall basketball player. Since Kitcher takes it to
be de dicto, 'Shorty' refers to Shorty in a and Kareem in w; thus, it is a necessary truth.
Kitcher's account is not only flawed, it does not help clarify the Kripkean examples.
Casullo concludes by investigating whether knowledge of one's actuality "stands in any
significant epistemic relationship" to the Kripkean examples of the contingent apriori (399).
Casullo has us consider ['If there is a shortest actual spy, then the shortest actual spy is a spy'](a).
He argues that this example by Kitcher is of the same sort as beliefs Kripke takes to be examples
of the contingent apriori. Kitcher does not say how this sort of example is related or similar to
knowledge of one's actuality. It cannot be the case that they are both examples of noninferential
a priori knowledge since that would require Kitcher's account to be correct and we have reason
to doubt its viability. However, Kitcher argues, more convincingly, that beliefs of the sort above
can be validly inferred from the necessary truth 'All A's are A's' and the knowledge of one's
actuality. We can produce a slew of contingent, a priori truths of the type 'All actual A's are
A's' in this same way.1
Casullo leads us through the chain of reasoning showing the connection between 'All
actual A's are A's' and knowledge of our own actuality (400). He has us consider 'Actual A's
are actual A's,' noting that it expresses a necessary truth in any world in which it is uttered.
Therefore, ['Actual A's are actual A's'](w*) is true. What do we need to move from this belief
to ['Actual A's are A's'](w*)?
The move cannot be immediate since the belief would not be true at all worlds; there
must be, therefore, some mediate knowledge to justify the inference. Casullo remarks that we
must know that, "the extension of 'A's' includes the extension of 'Actual A's' in the world I
inhabit." It is true apriori that if this is the actual world, then the extensions of 'A's' and
'Actual A's' will be the same. Therefore, if I know I am actual (i.e. that I inhabit the actual
world), I know the contingent, a priori truth ['Actual A's are A's'](w*).
1 Notice that this is exactly the sort of candidate that has seemed, thus far, to be a viable example of the contingent a
Casullo points out, however, that while the argument in the previous paragraphs is
sufficient, it is not necessary (401). If one takes an indexical view of actual,' 'actual' refers to
the world of utterance. Thus, the inference is justified since we need only know that w* is the
world of utterance. Casullo has thus shown that the contingent apriori is not inexorably linked
to knowledge of one's actuality.
Casullo concludes that we still have reason to doubt Kripke's claims about the contingent
apriori. Noting that his conclusion is similar to Donnellan's, Casullo writes that we have not
stated the content of the belief. He says that if we accept Kripke's semantic story about proper
names and definite descriptions, we may be able know apriori that some sentences express
contingent truths, but we do not know the truths they express. As I have already responded to
Donnellan along these lines, I shall not retrace my steps here.
While Casullo's paper seems to reveal that knowledge of one's actuality is not required
for knowledge of contingent, apriori truths, it does not give us any new reason to doubt our
current candidate. However, as Casullo notes, one might object that 'All actual geniuses are
geniuses' is unclear because the semantic role of 'actual' still has not been made explicit.
Though Donnellan's worries seem to not apply to examples involving 'actual', the surface
grammar may obscure rather than clarify the meaning of the sentence. We must, therefore,
analyze 'All actual geniuses are geniuses' in order to see whether it is an unproblematic
candidate for the contingent a priori.
Let's begin with the following analysis of 'All actual geniuses are geniuses':
(x)((x is a genius & x is in a) -* (x is a genius)),
where 'a' refers to the actual world. In this treatment, 'a' is a name that rigidly designates the
actual world. This immediately re-opens the door to Donnellan's objection. In order to know
the truth expressed by 'All actual geniuses are geniuses' we must know that our world is a; we
must know that 'a' refers to a. Can we know this apriori (or even at all)?
One way we can have knowledge is by description. For instance, I know that Plato was
the teacher of Aristotle and that the teacher of Aristotle was the student of Socrates, so I know
that Plato was the student of Socrates. Knowledge by description is in some sense inferential
knowledge. Can I know that this world is a through an inference of descriptions? No, because
any description we might offer would also be satisfied by another world if someone at that world
were to offer it. 'This world', 'the speaker's world', 'the world that exists', and other variants
will hold true at every possible world, though they won't express the same truth at each world.
What we need is a way to "pick out" our world from among all the possible worlds. This will
give us knowledge of a particular world, allowing us to know that 'a' refers to a by having
knowledge of a particular that also satisfies the description 'the actual world'.
Another way we can have knowledge is by acquaintance. For instance, I am acquainted
with my best friend and know things about him as a result of that acquaintance (e.g. that he is
married). Assuming we are acquainted with our world, there are two ways in which
acquaintance puts us in a privileged position to pick out "our world": (1) we are acquainted with
one and only one world our world, (2) no one in another world is acquainted with our world.
Thus, what we could not do with knowledge by description we can do with knowledge by
acquaintance know the particular world to which we want to refer.
So, is this world with which we are acquainted a? Is it the actual world? That depends,
in part, on whether the following accurately analyzes 'All actual geniuses are geniuses':
(x)((x is a genius and x is in my world) -* (x is a genius)),
where 'my world' is intended to act as an indexical, not as a description. I believe there are two
good reasons to think that this treatment of 'actual' is accurate. First, when we say 'actual
world' and 'my world' we certainly intend to refer to the same world, so the two phrases are
materially equivalent. Second, this analysis explains the reasons mentioned earlier why we
cannot have knowledge by description in this case. 'This world', 'the speaker's world', 'the
world that exists', and other variants will act as indexicals, referring to the world in which they
are uttered or considered. Thus, though they do not refer to a particular world when taken as
descriptions, as indexicals they will have the correct meaning at whatever world they are
evaluated while still referring (as considered at a particular world) to a particular world.
In what follows I will give a possible account of how knowledge by acquaintance can act
as the basis for us knowing apriori that 'a' refers to a (i.e. that our world is the actual world). I
will not insist that we are therefore acquainted with the actual world, but simply that we have the
knowledge of the actual world that we need.2
I will begin by giving a series of conditional premises that will lead us to this conclusion.
I will then provide support for the antecedent of the first premise and explanation for the
premises which may be found less than obvious. Thus, I will have proved that we can know a
priori that our world is the actual world, not simply that 'Our world is the actual world' expresses
a truth; this is equivalent to knowing that 'a' refers to a.
(1) If I think, then I exist.
(2) If I exist, then I am actual.
(3) If I am actual, then the world of which I am a part is the actual world.
2 It may, however, be the case that we are acquainted with our world.
(4) As a world is a complete state of affairs, my knowledge of myself either constitutes complete
knowledge of the world or partial knowledge of it.
(5) Either way, I know that there is an actual world and that my world is that world.
Perhaps (4) is in need of a brief explanation. My knowledge of myself would constitute
full knowledge of the world if solipsism is true and I have complete knowledge of myself (a
priori). Otherwise, I have only partial knowledge (apriori) of my world.
I will not bother to defend the claim that I know that I think. It is important, however, to
distinguish just what is meant by 'apriori' since it has been used in different ways. I will not
attempt to give a thorough account; I will simply give a rough definition that will capture what is
traditionally meant by the term and make explicit the feature relevant to this case.
Something can be known apriori iff it can be known without having any experience that
is not essential to acquiring the concepts involved (i.e. being able to use the words
The motivation behind this definition is that if we deny the cogitator reference to any and all
experiences, he will be unable to employ concepts since they require experience to acquire. We
maintain priority, however, by limiting the experiences involved to the minimum required to
acquire the concepts.
The Contingent A Priori and Knowledge of One's Actuality
It is important to notice that the link between the contingent apriori and knowledge of
one's actuality has once again been asserted, albeit in a way different from Kitcher's view. This
link holds if the above defense is required, but I do not believe it is. First, one might object that
while I have provided a seemingly viable analysis of 'is an actual genius', one must be
acquainted with the actual geniuses themselves if one is to even understand 'All actual geniuses
are geniuses', let alone believe it. As Donnellan suggests, if I do not know who the actual
geniuses are, how can I say of them that they are geniuses? I may know that 'All actual geniuses
are geniuses' expresses a truth, but I don't know the truth it expresses. It is similar to
'Fhqwhgads are Fhqwhgads'. Thus, what is needed is acquaintance with actual geniuses, not the
I think noting an important distinction between proper names and quantitative
expressions reveals the disanalogy between the current candidate and past ones. Proper names
are singular referring terms while quantitative expressions (definite descriptions, general
terms/expression, etc.) are denoting expressions; this distinction is based on Russell's
observations. I will not go into the arguments for and against the distinction between referring
and denoting, other than noting the supporting intuition that the connection between a proper
name and its referent is more direct than is the connection between a definite description and the
thing it describes.3 A proper name just is a title or label for an individual thing, while a
quantitative expression simply picks out the thing which uniquely satisfies it. How does this
distinction help us?
Consider 'Newman-1 is the first child born in the 21st century'. The best way to
represent 'Newman-1' would be with a constant rather than a variable, a formalization of the
where 'F' means is thefirst child born in the 21st century and 'c' refers to Newman-1. Since the
formalization requires a constant, knowledge of the proposition requires knowing that c is the
first child born in the 21st century. It requires knowledge of Newman-1 than cannot be gained
just by the stipulation that was its origin, or by description. It requires acquaintance.
3 As evidence, reflect on the difference between 'The author of Huckleberry Finn is Mark Twain' and 'Samuel
Clemens is Mark Twain', or 'Mark Twain is Mark Twain'.
Now consider 'All Martians are green', which could be represented by:
(Vx)(Mx -- Gx).
Here we have a universally-quantified variable rather than a constant. Thus, we do not need to
be acquainted with Martians to understand the proposition. We don't even to know whether such
beings exist to understand the proposition. So, while our previous candidates may have been
problematic, we have no reason to think that of 'All actual geniuses are geniuses'.
Two-Dimensional Logic and the Contingent A Priori
In the paper by Bostock discussed earlier, he mentions a paper by Martin Davies and
Lloyd Humberstone in which a two-dimensional modal logic is used to handle the modality of
'actual' and its variants. In what follows I want to look at the system, partly to understand our
current candidate better and partly to see whether their system attempts to undermine, in any way
not yet accounted for, the viability of 'All actual geniuses are geniuses' as an example of the
contingent a priori.
In "Reference, Contingency, and the Two-Dimensional Framework," Davies begins by
characterizing what he calls 'the two-dimensional framework'.4 There are some sentences, like
'It is possible that everything that is actually red should have been shiny,' that resist formulation
using traditional modal logic. The usual way of handling these sorts of sentences is to introduce
an actuality operator, 'A' (84). Davies writes,
"In terms of possible-worlds model-theoretic semantics for the modal language, a
sentence 'As' is true with respect to a possible world, w, just in case the embedded
sentence s is true with respect to the model's designated or 'actual' world, w*.
One result of this, which was noted earlier in this paper, is that if 'As' is true at all it is
necessarily true. This seems hard to reconcile, however, with the fact that what is actually the
4 Philosophical Studies, 118, 2004 (pgs. 83-131)
case is (at least mostly) a contingent matter. This leads Davies (and Humberstone) to suggest
two sorts of necessity to clarify the difference, and an additional modal operator to help
characterize the two types of necessity.
Davies and Humberstone suggest 'T', a "fixedly"-operator (85). While 'A' increases our
modal language's ability to express sentences by allowing us to vary our world of evaluation, wj,
'F' allows us to vary what world plays the role of the actual world, wi. Davies writes,
"A sentence 'os' is true with respect to a world wj with world wi playing the role of the
actual world just in case, for every world w, the embedded sentence is true with respect to
world w, with wi still playing the role of the actual world. A sentence 'Fs' is true with
respect to a world wj with world wi playing the role of the actual world just in case, for
every world w, the embedded sentence s is true with respect to wj, but now with w
playing the role of the actual world" (85-86).
How this operator gives us two different notions of necessity can be seen by considering TAx'.
If s does not contain A, Ax' is equivalent to 'os'. If, however, x is of the form 'As', where s is
some contingent truth, then a difference is revealed. While 'ox' is true, 'TAx' is false since it is
equivalent to As' and thus to 'os'. As a result we have two different sorts of necessity,
allowing us to maintain the intuition behind the semantics of 'actually' and the intuition that it is
a contingent truth that the world that is in fact the actual world is the actual world.
Thus the two-dimensionality of this system of modal logic is revealed (87). Before 'TF
the only world we needed to consider when evaluating a sentence was wj; the logic was, thus,
one-dimensional. Now, however, we must also consider wi since it can vary; thus, our logic is
two-dimensional. While Davies goes into more detail about this and addresses many related
tangents, I shall not go into those considerations here. Rather, we shall conclude our look at his
paper by examining his comments on the contingent a priori.
Before going into Davies' comments we shall give Evans' characterizations of superficial
(corresponding to 'o') deep (corresponding to '~F) necessity and contingency, which Davies
summarizes and uses in his comments (93-94).5 Superficial contingency is a property of
sentences, and depends on how the sentence embeds under the scope of 'o' or '0'. Davies lists
three ideas associated with superficial contingency to help us understand the notion: truth with
respect to worlds, purely internal to semantic theory, and properties of modal sentences (95). On
the other hand, three ideas associated with deep contingency are: truth in worlds (being made
true), not purely internal to semantic theory, and modal properties of sentences. Davies inserts
the following quotation by Evans,
"If a deeply contingent statement is true, there will exist some state of affairs of which we
can say both that had it not existed the sentence would not have been true, and that it
might not have existed. The truth of the sentence will thus depend upon some contingent
feature of reality."
Superficial contingency and necessity are simply the negation of deep contingency and necessity.
Davies spend some time exploring and defending the notion of 'truth in a world' and
showing that his D-necessity and Evans' deep necessity are one and the same even though the
former is put in terms of a modal operator, T, and the latter is not. However, I wish to move on
to Davies' application of two-dimensional modal logic to alleged cases of the contingent a priori.
Davies begins by expounding Evans' notion of epistemic equivalence (99). Essentially,
two statements are epistemically equivalent iff they have the same content and, if both are
5 Davies calls superficial necessity 'H-necessity' and deep necessity 'D-necessity' since 'V' is truth along the
horizontal and 'F' is truth along the diagonal in a two-dimensional system.
understood, one cannot believe one and disbelieve the other. One example of this which is
important for our investigation is the epistemic equivalence of 's' and 'As'. One result of this is
that we are in a position to know 'As -* s' apriori. Davies' example is that if someone
understand the notion of actuality, he is in a position to know apriori 'if the earth moves, the
Earth actually moves'.
Davies notes several things about this example. First, s can only be known aposteriori,
so As can likewise only be known aposteriori, though 'As <- s' is known apriori. Second,
while 'o(As)' is true, 'FA(As)' is false (100). 'As' is superficially necessary and deeply
contingent. 'As -* s', however, has the opposite properties; it is superficially contingent and
deeply necessary. Thus, our example, and those of the same form, seems to show a correlation
between a priority and deep necessity, though nothing about two-dimensional modal logicper se
Davies now gets to the heart of the matter, investigating whether there are deeply
contingent, apriori truths. He says there are two problems: 1) there are some noteworthy
counter-examples to the claim that a priority and deep necessity coincide, and 2) while intuitions
seem not to favor the contingent apriori, it is very difficult to form an argument from them. We
will conclude our look at Davies' paper by considering each of these in turn.
Davies gives the example of the belief that one has hands (101). While it is certainly
possible that I am simply a brain in a vat, I am entitled to ignore this possibility so long as there
is no evidence in support of it, and am justified in believing I have hands. Similarly, Davies
notes that when doing a mathematical proof, one is entitled to ignore the possibility (so long as
there is no supporting evidence) that one is suffering from a massive memory loss and, thus,
cannot be sure the proof is being done correctly. In these cases, Davies alleges, one has apriori
beliefs for which defeaters could arise (i.e. certain experiences would reveal them to be false, or
at least unjustified) (102). Thus, we have certain a priori beliefs about contingent states of
For my part, I do not see how Davies' examples can properly be considered apriori. He
says that I am entitled to believe these things so long as there is no contradictory evidence, but
how do I know apriori whether there is contradictory evidence or not? I need to know there is
no contradictory evidence for my belief to be justified, but I can only know this aposteriori.
Furthermore, the belief that I have hands was formed aposteriori; I had to see my, or perhaps
someone else's, hands. It seems that if anything here is apriori it is the epistemic warrant 'I am
entitled to believe x so long as there is no evidence to the contrary', where 'x' is a variable
ranging over a certain type of belief. If this warrant, or something like it, is true at all it is
necessarily true, so Davies does not bolster the case for the contingent apriori.
Davies gives the following reduction adabsurdum. Suppose there is a truth, s, which is
both able to be known apriori and deeply contingent. Now, s is true iff a certain state of affairs,
S, obtains. However, S may or may not obtain; there is no guarantee that it will obtain (i.e. it is
not a deeply necessary truth). Davies says that this may appear to be a contradiction since s
would as a guarantee of S, but says this is a confusion between the modal and the epistemic
notion of 'guarantee'.6 What is needed is a step showing that if S is not modally guaranteed than
it cannot be epistemically guaranteed (by s or any other proposition), but Davies notes how
similar this is to the assertion that if something is apriori, then it is deeply necessary the very
thing we wished to prove (103).
6 Though Davies does not say exactly what this amounts to, he says s acts as an epistemic guarantee for S, but S is
not modally guaranteed.
So in the end, while Davies admits that there are strong intuitions (at least for some)
against the possibility of contingent, apriori truths, he has given us no reason to doubt the
possibility, nor does he claim to. Note that while Bostock's claims about the impossibility of
(philosophically interesting) contingent, apriori were much stronger than Davies, his use of two-
dimensional modal logic did not present any problems more substantial than those Davies raises.
Casullo's paper helped us understand the problems with the Kripkean examples and how
our current candidate does not fall prey to those problems. It did, however, reveal the need for
an analysis of our current candidate. We have seen that a viable analysis of 'is an actual genius'
can be given that not only allays concerns about the meaning of 'All actual geniuses are
geniuses', but also separates the question of the contingent apriori from questions about the
knowledge of one's actuality. The two-dimensional modal logic of Davies and Humberstone has
helped us understand our current candidate by revealing how the rigid designation accomplished
by 'actual' affects the modal status of the proposition. We can also conclude that we have seen
no reason to doubt that 'All actual geniuses are geniuses' is both contingent and apriori, and that
(bar some refutation being given) we should accept it as such.
A NON-INDEXICAL CANDIDATE FOR THE CONTINGENT A PRIOR
It is interesting that all of the candidates considered thus far have relied on indexicality in
some way. In some cases this is clear, and in others the indexicality would be revealed if we
were to analyze the proposition in question. One might be tempted to think that the contingent a
priori, if there is any such thing, depended upon indexicality. Timothy Williamson, however,
wishes to separate indexicality from questions about the contingent a priori by providing and
defending an indexical-free, contingent, apriori truth.
Williamson and Oppy's Conversation
Williamson begins by suggesting the following:
(1) There is at least one believer.1
Williamson notes that 'is' must be read tenselessly if the proposition is to be indexical-free. It is
clear, setting aside for the time being the worry that God is a necessarily existing believer, that (1)
is contingent. However, does one know (1) apriori?
Williamson writes, "Now since it is impossible to believe (1) falsely, it seems that one
can know (1) a priori" (114). Does this condition ensure a priority? At the end of his paper
Williamson briefly discusses the problem of providing a formal definition of 'a priori'that lines
up with our understanding of how the term should be employed. He provides a couple of
examples, shows how they are problematic, and concludes by saying that while one might
question whether his condition is necessary, it is certainly sufficient. While he does not say why,
I believe it is because the nature of a posteriori truths would seem to require that it be possible
that they be believed falsely. Thus, anything which is impossible to believe falsely must be a
priori. I see no problem with this, and agree that Williamson's condition is at least sufficient.
1 Analysis, 46.3, June 1986 (pgs. 113-117). The C. ',i,,A,. rA Priori: HasIt, i, lr,,l. To Do With Indexicals?
But now the worry is that one cannot know (1) without the use of indexicals. While (1)
does not contain them, it is arguable that one cannot know that (1) is true without knowing:
(2) I am a believer.
If this is true we are presented with a dilemma. We can either admit that indexicality is, even in
this case, required for something to be a contingent, a priori truth, or we can say that (2), and
thus (1) as well, is not a priori at all because it is based on introspection. Therefore, we either
have yet another indexical-based example of the contingent apriori, or no example of it at all.
Williamson responds by arguing that one can know (1) a priori without the use of (2).
He begins by providing the following method of forming beliefs:
(M) Given a valid deduction from the premise that someone believes that P to the conclusion
that P, believe that P.
The set of beliefs formed according to this principle, or belief-forming mechanism, would
include only true beliefs (that is, (M) is infallible). Since this fact can be known a priori it
follows that any knowledge produced in accordance with (M) is itself known apriori so long as
the "input" is knowable apriori. If we replace P with (1), the deduction from,
(3) 'Someone believes that there is at least one believer',
to (1) is clearly valid (115). Looking at (M), we see that the deduction from (3) to (1) is the
input. We can certainly know that the deduction is valid apriori, so we have an apriori warrant
for believing (1), a contingent truth.
(M) can also be used to generate other indexical-free, contingent, a priori truths. For
instance, 'There is at least one thing which exists' is certainly contingent, and by (M) we can
know it apriori without reference to the 'I exist'.
Williamson concludes his argument by addressing the objection that God is a necessarily
existing believer, thus making (1) necessary rather than contingent. We can modify (1), getting:
(1') There is at least one fallible believer.
The deduction for (M) goes as follows. If someone believes that (1'), this being is either fallible
or infallible. If fallible, then (1') follows trivially. If infallible, then the being's belief that (1')
must be true. So, we can know (1') a priori, and it is certainly contingent since there are possible
worlds in which God, an infallible believer, is the only believer that exists.
In "Williamson and the Contingent A Priori" Graham Oppy denies that Williamson has
adequately supported his conclusion.2 Oppy begins by arguing that the way (M) is formed hides
the fact that it actually contains an indexical (189). Rather than putting things in the imperative
(i.e. 'believe p'), Oppy suggests two ways one could use (M).
(M') If from the fact that I believe that P it follows that P is true, then I shall have a true
belief if I believe that P.
(M") If from the fact that someone believes that P it follows that P is true, then anyone can
form a true belief by believing that P.
(M') is clearly problematic because of the indexical. Oppy argues that because (M") is
universally quantified and (1) is existentially quantified, while we can know (M") apriori, it has
not been shown that we can know (1) apriori.
Before continuing with Oppy's paper let's look more closely at his claim. If things are as
he says and one is trying to infer an existential from only a universal, then his objection would be
valid because of the lack of existential import. However, this is not the case. We are inferring (1)
not from (M), but from 'Someone believes that (1)', another existentially quantified statement.
On the basis of this deduction the universally quantified 'Anyone can form a true belief by
believing that (1)' is made by way of (M), but this suffers no problems with existential import.
Oppy's objection, therefore, is unsubstantiated.
2 Analysis, 47.3, October 1987 (pgs. 188-193)
Oppy has misunderstood (M). (M") is intended to be a proposition in a deduction, while
(M) is a belief-forming mechanism, not a proposition. Furthermore, it is (in some ways) similar
to the reading that Oppy anticipates Williamson would use in replying to his objection (190).
Oppy formulates this reading as:
(APK) Bxp -p p
He notes that it is similar to modusponens except that it is only truth-preserving when used. He
then argues that even on this reading of (M) the inference to (1) is not justified.
Oppy has us suppose that there is a computer that exists in a world in which there are no
subjects (i.e. no one who has beliefs). If this computer operates in accordance with modus
ponens it will produce only true results, but this is not so for APK. The machine will generate
'There is at least one believer', which is false. Therefore, any inference based on APK would
not be valid.
Oppy admits that this objection is somewhat questionable, but I believe that it is
hopelessly flawed. Let's remember that the purpose of (M) is to produce truths that are
impossible to believe falsely. So, does the thought experiment above disprove Williamson's
claim? No. The computer does not have beliefs at all, so it certainly does not have any false
beliefs. If it did have beliefs, then (1) would be true. Let us, therefore, move on to another
objection Oppy lays out.
Consider a subject with limited self-awareness (i.e. who can competently use words like
'believer' without knowing that the terms apply to him) (192).3 Now, as far as he knows his
world has no subjects. So, what will keep him from concluding, via APK, that some world other
than his own, which happens to be subject-less, has at least one believer? Oppy believes that it is
3 This presumably helps to avoid sneaking in knowledge based on indexicals.
a hidden indexical, linking the conclusion to the world of utterance (i.e. the actual world). After
all, someone wishing to use APK can only use it validly in her world, though there is no explicit
reference to her world or her as the believer which satisfies the condition.
Oppy says that one might object that we "simply take it for granted" that we are talking
about the world of utterance when there are no explicit operators or indexical elements. He says,
however, that this is irrelevant since it is a pragmatic concern; logically, we must take such
indexicality into account. Oppy suggests the following representation of APK, and thus
formalization of (M):
where (0) is an operator which fixes the world at which p, and thus the whole argument, is to be
evaluated as the world of utterance. Therefore, Oppy concludes, Williamson's argument
provides no reason to think that one can separate the contingent apriori from indexicality (193).
In "The Contingent A Priori: A Reply", Williamson responds to Oppy's objection.4
Williamson argues that APK is not an accurate representation of (M), and that APK itself is
problematic (219). Williamson writes, "[I]f p is false and believed by no one, APK will take you
from the vacuously true supposition Bxp -* p to the false supposition p." Williamson says that
the following would be more similar to (M) and would also produce indexical-free examples of
the contingent a priori:
(MAPK) Given the belief that if someone believes that P then P, believe that P.
He remarks that both (M) and (MAPK) should be understood, not as inference rules, but as
"something like abstract mechanisms for generating beliefs."
4 Analysis, 48.4, Oct. 1988 (pgs. 218-221)
There are two problems with Williamson's remarks here. First, he is operating with a
rather radical notion of the a priori. As Fitch points out, if our experience is used to justify our
belief in the proposition, rather than simply help us acquire the requisite concepts, then our
knowledge should not be considered a priori. Thus, knowledge produced by (M) and (MAPK)
is not a priori.5 (MAPK) is suspect for an additional reason. The belief 'If someone believes
that P then P' could be false. For instance, I could believe that if someone believes that pigs can
fly then pigs can fly. In this case, (MAPK) would generate the false 'Pigs can fly'. Thus both
(M) and (MAPK) fail to be truth-preserving.
One might also worry that the imperative mood of (M) and (MAPK) is masking some
sort of indexicality (that is, that it is a trick for hiding the indexicality that would be clear if (M)
or (MAPK) were put in propositional form as part of a deduction. However, since (M) and
(MAPK) are belief-forming mechanisms rather than steps in deductions, Williamson's examples
do not make use of indexicals in the way Oppy believes it does. Therefore, while there may be
reasons to be worried about Williamson's account, they are by no means decisive rebuttals of it.
Williamson notes that Oppy's indexicalized APK suffers from the same problems as the
version without indexicals and, as a result, fails to be truth preserving (220). If, however, one
tries to show that (MAPK) fails to be truth-preserving in the case of the subject-less world this
attempt will fail because the input, 'If someone believes that in w there is at least one believer
then in w there is at least one believer,' is itself false.
Williamson addresses a possible concern that Oppy mentioned but did not dwell on.
What of the objection that it is questionable whether someone who does not know that he is a
believer understands 'There is at least one believer'? Williamson writes,
5 It is possible that Williamson's thoughts about the nature of the a priori stem from his Reliablist position.
However, since the other authors we've discussed all seem to accept the more traditional notion of the a priori, it is
outside the scope of this paper to investigate the case for Williamson's notion.
"What should also be noted is that even if a negative answer were correct, it would
constitute no objection to my argument. For it would not show my a priori knowledge
(with understanding) of the contingent truth that there is at least one believer to be
indexical-dependent in any sense not applicable to my a priori knowledge of the
necessary truth that all believers are believers."
One could, of course, argue that we cannot know 'All believers are believers' apriori because of
the sort of experience necessary to acquire the concept 'a believer', but this seems implausible,
and would undermine many of the traditional examples of the a priori.
While Oppy's objections to Williamson's examples are off the mark, there are reasons to
doubt that his examples give us examples of the contingent apriori. Both (M) and (MAPK) fail
to be truth-preserving in worlds without subjects, albeit for different reasons. Furthermore, we
could at least question whether the examples are really indexical-free because of the imperative
mood. It would seem, therefore, that Williamson does not give us a reason to believe that the
contingent apriori can be separated from indexicality.
In what follows I want to briefly review each of the sort of candidates considered,
summarizing why they are or are not true examples of the contingent apriori. I will also point to
the ways in which the failure of the earlier candidates shaped the formation of the later ones.
The incorrigibilityy" examples were dismissed rather quickly by Fitch. The problem with
this class of examples is that they do not qualify as apriori truths. They all depend on
experience in order to provide justification for belief in the proposition in question. While even
legitimate examples of the a priori depend on experience, it is to acquire the concepts involved
rather than justify belief in the proposition.
The Kripkean examples (the "Standard Meter Bar", "Neptune", etc.) fail to be examples
of the contingent apriori. The main problem with this sort of example, revealed by Donnellan
and Fitch, is one of the results of the act of rigid designation (that is, the stipulative introduction
of the name involved). The name lacks sense, so it is impossible to know what the sentence
means without having knowledge of a particular (the referent); thus, we cannot know the
proposition a priori. The upshot of this sort of candidate is if we could know the truth the
proposition expresses, the rigid designation allows us to say things about a particular that are true
apriori in this world and false in some possible worlds.
To avoid the problems with the Kripkean examples, some have suggested examples using
'actual' and its variants, which allow for rigid designation without robbing any of the terms of
their sense. These examples hang on a distinction between the world that determines the
extension of the predicate or truth of the proposition and the world at which the proposition is to
be evaluated. While Bostock tries to show that the utilization of this distinction requires changes
in our system of modal logic and a dissociation of the notion of 'necessity' from "truth in all
possible worlds", I show that his assertions are questionable at best. In addition, whatever merit
his observations have applies only to wide-scope uses of 'A', leading me to suggest a candidate
that relies on a narrow-scope use of 'A': 'All actual geniuses are geniuses'. I also dismiss
Bostock's 'If there is exactly one object that falls under the predicate, then it is necessary that
there is no more than one'. To understand the sentence one must understand the rigidified
predicate in question. However, one cannot understand it apriori because, like the Kripkean
examples, it lacks sense, and Donnellan and Fitch show us it would take knowledge of the
particulars involved, something we can only have aposteriori, to understand the predicate.
I then re-consider objections by Donnellan, Kitcher and Bostock, demonstrating that 'All
actual geniuses are geniuses' can be known apriori and that knowledge of this truth produces
more than just linguistic knowledge. In response to the objection that perhaps the meaning of the
sentence is still unclear, I provide a viable analysis that utilizes a distinction between knowledge
by description and knowledge by acquaintance. This distinction helps reveal how the current
candidate differs from the former ones. The previous candidates involved proper names. Since
proper names are singular referring terms, more is required to have knowledge of the referent of
a proper name than is required for a definite description, a denoting expression. This helped us
see more clearly why the previous candidates failed where the current candidate succeeds.
The paper by Davies, while not giving us a reason to doubt our current candidate, helped
us under the trick being accomplished by 'actual'. By allowing us to vary the world of
evaluation, 'actual' and its variants increases the power of our modal system, allowing us to
assert and evaluate things we could not before. It is the increased power that allows for
examples of the contingent apriori to be produced.
Williamson, wanting to separate the question of the contingent a priori from indexicality,
sought to offer an indexical-free, contingent, apriori truth. However, as we saw, his examples
failed, strengthening the case that it is the special sort of indexicality of 'actual' that allows for
examples of the contingent apriori.
LIST OF REFERENCES
Bostock, David. "Necessary Truth and A Priori Truth." Mind, New Series, Vol. 97, No. 387, Jul.,
1988 (pp. 343-379)
Davies, Martin. "Reference, Contingency, and the Two-Dimensional Framework." Philosophical
Studies, 118, 2004 (pp. 83-131)
Donnellan, Keith. "The Contingent A Priori and Rigid Designators." Contemporary Perspectives
in the Philosophy of Language, eds. French, Uehling, and Wettstein. University of
Minnesota Press, 1980 (pp. 45-60)
Fitch, G.W. "Are There Contingent A Priori Truths." The Journal of CriticalAnalysis vol. 6,
no. 4, Jan./Apr. 1977 (pp.119 123)
Kitcher, Philip. "A Priority and Necessity," from Australasian Journal ofPhilosophy, 58, 1980
Kripke, Saul. Naming and Necessity. Harvard University Press. 1980
Oppy, Graham."Williamson and the Contingent A Priori."Analysis, 47.3, Oct. 1987
Williamson, Timothy. "The Contingent A Priori: A Reply."Analysis, 48.4, Oct. 1988
Williamson, Timothy. "The Contingent A Priori: Has It Anything To Do With Indexicals?"
Analysis, 46.3, June 1986 (pp. 113-117).
Daniel Carter McCain was born on July 16, 1982 to Daniel Ross McCain and Doris
Carter McCain. He was raised in Jacksonville, Florida and graduated from Englewood High
School in 2000. He also attained the rank of Eagle Scout in the Boy Scouts of America. Carter,
as he prefers being called, attended the University of South Florida and then the University of
North Florida, where he graduated cum laude with a B.A. in philosophy in 2004.
After graduating from the University of Florida in 2006 with an M.A. in philosophy, he
will marry Karen Garito and transition into professional ministry with the Graduate and Faculty
Ministries branch of InterVarsity Christian Fellowship. His long-term goals are to attend
Seminary, attain an M.Div., and to eventually serve in a pastoral role at a local church.