Pluralistic Skepticism: Advertisement for Speech Act Pluralism
University of Oslo
Forthcoming in: Philosophical Perspectives, 19, Epistemology, 2005
Even though the lines of thought that support skepticism are extremely compelling,
we're inclined to look for ways of blocking them because it appears to be an impossible
view to accept, both for intellectual and practical reasons. One goal of this paper is to
show that when skepticism is packaged right, it has few problematic implications (or at
least fewer than is often assumed). It is, for example, compatible with all the following
claims (when these are correctly interpreted):
• We can say something true by uttering sentences of the form "S knows that p";
that is, our ordinary, reflective, intuitions about the truth-value of knowledge
attributions can be respected.
• The beliefs expressed by utterances of sentences of the form "S knows that p" can
• Knowledge is the norm of assertion.
Skepticism is compatible with these claims when combined with a view I call Speech Act
Pluralism. Speech Act Pluralism is a theory about the content of speech acts -- it's a
general theory about the relationship between semantic content and speech act content.
It's motivated by considerations that have nothing specifically to do with skepticism or
epistemology. In other words, this is not an ad hoc solution to problems involved in
accepting skepticism. It is, rather, a theory that anyone who thinks about the nature of
linguistic content should adopt. It just so happens that it lends support to skepticism.
A wide range of philosophical arguments in all areas of philosophy appeal to the
intuitions that competent speakers have about the content of sentences, what's said by
utterances of sentences, and the truth-conditions of such utterances. Proponents of Speech
Act Pluralism claim that many such arguments are fundamentally flawed because they
are based on mistaken assumptions about the nature of speech act content (i.e. the nature
of what's said, asserted, claimed, etc by utterances) and the relation between speech act
content and semantic content. Much work done in epistemology (especially the debates
that focus on the semantics for "know") provides a good illustration of this contention.
Many of the central arguments against skepticism appeal to intuitions we allegedly have
about what's said and asserted by utterances of knowledge attributions. If Speech Act
Pluralism is correct, these arguments uniformly fail.
There's a lot that I'm not doing. In particular:
• I don't provide new arguments for skepticism -- I take familiar skeptical
arguments at face value, i.e. as showing that skepticism is true. I then defend that
view from various objections. Of course, indirectly, this is an argument for
skepticism because if I'm right, some of the central obstacles to accepting
skepticism are removed.
• I don't present direct arguments against alternative semantics for "know". In
particular, I don’t present all the data that show that "know" is not a context
sensitive term. 1 However, indirectly I provide an argument against contextualism
because if I'm right, all the evidence that's alleged to support contextualism is
better explained by my version of skepticism.
• I don't defend Speech Act Pluralism. I present the view, provide some
illustrations, and refer the reader to the various places where the arguments for it
are developed in greater detail. The goal here is to illustrate the philosophical
significance and usefulness of Speech Act Pluralism, not to present all the
arguments for it.
The paper has three parts. Part One presents Pluralistic Skepticism in more detail.
Part Two shows how it can account for speakers intuitions about sentences containing
"know". Part Three is about the relationship between Pluralistic Skepticism and the
knowledge account of assertion.
Pluralistic Skepticism (PS) has three central components:
a. Semantic invariantism: the view that the semantic value of "know" is invariant
between contexts of utterance.
b. Skepticism: the view that it is extremely hard to know anything. The semantic
content of utterances of sentences of the form "S knows that p" are almost never
true. We don't know most of what we take ourselves to know. We stand in the
knowledge relation to few, if any, propositions.
c. Speech Act Pluralism: the view that in uttering a sentence one (literally) asserts
indefinitely many propositions (only one of which is the proposition semantically
expressed.) As a result, it doesn't follow from a. and b. that one can't (literally) say
something true by uttering a sentence of the form "A knows that p".
Before I present this view in more detail, I'll give a brief overview of how it is related to
some other positions one might hold about the semantics for "know".
First, PS is opposed to traditional versions of skepticism2. According to this
view, skeptical arguments reveal that what's literally said by utterances of sentences of
the form "A knows that p" is always (or almost always) false. Our intuitions to the effect
that such utterances are true are explained away either as a mistake (i.e. as the result of
being mistaken about how hard it is to obtain knowledge) or as the result of confusing
warranted assertability with true assertion (i.e. confusing the truth of an assertion with its
It is the (c) component of PS that distinguishes PS from this kind of skepticism.
According to PS, our intuitions about the truth-value of utterances of sentences (relative
to circumstances of evaluations) are intuitions about what's saliently asserted by those
utterances. When what's saliently asserted by an utterance of "S knows that p" is not the
proposition semantically expressed (and typically it isn't), we will have the correct
intuition that this utterance said something true (even though the semantic content is
Second, PS is opposed to contextualism about "know".3 According to
contextualists, the semantic value of "know" shifts from one context of utterance to
another -- know belongs in the same semantic category as "she", "that", and "you". The
primary motivation for contextualism is the kinds of intuitions we have about variability
of what's said (and the truth-values of what's said) by utterances of knowledge
attributions. According to PS this is best explained, not as semantic variability, but by
variability in what's saliently asserted from one context of utterance to another (and from
one context of interpretation to another.)
Third, PS is opposed to various versions of non-skeptical invariantism4. The
non-skeptical invariantists agree with PS that the semantic value of "know" is invariant
between contexts of utterance. They differ from PS'ists in claiming that the semantic
content of utterances of "S knows that p" is true (in (most of) the cases where we
intuitively think it is true). The non-skeptical invariantists try to find a stable semantic
value for "know" that doesn't make all knowledge attributions false. They do that (at least
in part) because they try to respect our intuitions about what speakers say (and the truth-
value of what they say) when they utter sentences containing "know". Again, this is,
according to PS, a mistaken strategy -- it is an attempt to give a semantic account of
variability in speech act content.5
Note that there is something the last two opponents of PS agree on: there's data
about our intuitions about what's said by utterances of knowledge attributions that it
seems difficult for a skeptic to account for. That's one reason why both the contextualist
and non-skeptical invariantist go to great lengths fixing up the semantic value of 'know'
(the former by making it context sensitive, the latter by making it, for example, subject
sensitive.). A central thesis of this paper is that the contextualist and the non-skeptical
invariantist make the same mistake: they try to account for variability in speech act
content by tinkering with the semantics.
Before showing how PS deals with the kind of data that makes contextualists and
non-skeptical invariantists opposed to skepticism, I need to say a bit more about two
components the version of skepticism I'm defending: the skepticism part and the speech
act pluralism part.
The arguments for skepticism are familiar, but two points are important to
emphasize in order to prepare the defense of PS: the effects of skeptical arguments and
the relationship between skepticism and semantic competence. I discuss these in turn.
The Effects of Arguments for Skepticism
One way to think about skepticism is as the view that the semantic value of
"know" is such that extremely high standards must be met in order for the proposition
semantically expressed by a positive knowledge ascription to be true. According to this
view, the propositions semantically expressed by all (or almost all6) utterances of
sentences of the form "A knows that p" are false. The arguments for this are old and
familiar; they typically involve evil demons, brains in vats, or so-called lottery
propositions. Here's an illustration of the latter:
I think I know that there's a computer in my office at the University of Oslo.
However, when I ask myself (sitting at home) whether I know that no one has
broken into my office the last 20 minutes and stolen my computer, I'm inclined to
say that I don't know. UiO is a relatively safe campus, but burglaries do happen,
and I can't rule out that today I was the unlucky one. This leads me to think that I
don't stand in the knowledge relation to the proposition that no one has broken
into my office in the last 20 min and stolen my computer (call this proposition 'the
lottery proposition') The proposition that I know that there's a computer in my
office, entails the lottery proposition. Realization that I don't know the lottery
proposition leads me towards the conclusion that I don't know what I thought I
knew, i.e. that I don't know that there's a computer in my office.
Two points about this kind of argument are important for what follows:
a. Skeptical arguments of this kind intuitively generalize in three important ways:
They make me conclude not just that I don't know that there's a computer
in my office (right now), but more generally, that I know very little of
what I though I knew. As Hawthorne puts it:
"These considerations generate a powerful pressure towards a skepticism
that claims that we know little of what we ordinarily claim to know. For
when confronted with the data, we philosophers feel a strong inclination to
stick to our judgment about the lottery proposition and retract our original
judgment about the ordinary proposition" (Hawthorne, (2004), p. 6)7
They make me draw conclusions about my epistemic state at other times. I
don't just conclude that I know very little now -- I conclude that I never
Finally, they make me conclude that others who claim that I have
knowledge are wrong. I don't draw conclusions just about my own self-
ascriptions -- anyone who says about me that I know that p (where p is one
the propositions I now have figured out that I don't know) is wrong.
b. Skeptical arguments are accompanied by a sense of discovery -- by a sense of
having understood something new about our epistemic condition. A theory of
knowledge should account for this sense of discovery that accompanies skeptical
arguments. Below I argue (what might seem obvious) that no non-skeptic theory
can do that.
I take (a) and (b) to be data about the effects skeptical arguments have on people
when they first encounter them. An adequate theory should explain why those arguments
have those effects and that can be done in one of two way: either explain these reactions
away as some sort of confusion, or take them at face value and adjust your theory of
knowledge accordingly. The skeptic pursues the latter option.
Skepticism and Understanding of "know"
According to PS, non-philosophical speakers don’t know that skepticism is true --
it is not a necessary condition on understanding "know" that speakers know that the
semantic content of utterances of sentences containing "know" typically are false. That's
something they find out by thinking about philosophical arguments. This is connected to
point (b) above: Skeptical arguments wouldn't be accompanied by a sense of discovery if
linguistic competence with "know" required knowledge of the truth of skepticism.
This is not a surprising feature of the verb "know". For any term F, there will be
important truths of the form something is F just in case G, knowledge of which is not
required for understanding F. An analogy might clarify this point: There are many truths
of the form A loves B only if … knowledge of which are not required in order to
understand and be a competent user of the term "love". No sensible semanticist would
assume that the semantics for English must include all such truths -- the semantics for
English will not include a theory of love (or, for that matter, of power or justice.) And
that's fortunate, for if it did, none of us would be competent English speakers. The same
goes for "know": you don't need to know a theory of knowledge in order to understand
English -- in particular, you don’t need to know that skepticism is true.8 9
Speech Act Pluralism (SPAP)
The most distinctive part of PS (what distinguishes it from other versions of
skepticism) is that it incorporates Speech Act Pluralism (SPAP). SPAP is not a view that
most philosophers are familiar with, so I need to go into some detail to present its main
components. As mentioned above I will not present detailed arguments for the view
here.10 In stead, I'll present three components of SPAP that will be of importance in
explaining how PS accounts for allegedly anti-skeptical data.
SPAP Elaboration 1: Pluralism
SPAP is the view that any utterance of a sentence S in a context C says (asserts,
claims) many propositions other than the semantic content of S relative to C. An
important corollary is that it is not the case that if an utterance of S in a context C says
(asserts, claims) that p, then p is the semantic content of S relative to C.
Illustration #1: The Dresser: (from Cappelen and Lepore (1997): Imagine an
utterance of (3).
3. A: At around 11 p.m., I put on a white shirt, a blue suit, dark
socks and my brown Bruno Magli shoes. I then got into a waiting
limousine and drove off into heavy traffic to the airport, where I just
made my midnight flight to Chicago.
According to SPAP, (4)-(6) are all true descriptions of what's said by an
utterance of (3) (note that 4-6 are all different propositions):
4. A said that he dressed around 11 p.m., went to the airport and
took the midnight flight to Chicago.
5. A said that he dressed before he went to the airport.
6. A said that he put on some really fancy shoes before he went to
The extent to which 4-6 will seem natural will depend on the circumstance of
the report, so arguments for SPAP are accompanied by small stories that
describe the context for the report. Having argued that these reports are
literally true (not just appropriate or warranted), SPAP proponents conclude
that in uttering 3. A (literally) said the complement clause of 4-6. And that's
just a tiny sample of what was said in uttering 3.
Illustration #2: The Terrorist: Here's a similar example from Scott Soames
"A terrorist has planted a small nuclear device in a crowded stadium
downtown. There is no time to evacuate the building or the surrounding
area. In speaking to the negotiator, he says "I will detonate the bomb if my
demands are not met," knowing that it is obvious that if he does so,
thousands of people will die, and intending to communicate precisely that.
The negotiator reports to his superior that the terrorist said that he will kill
thousands of people if his demands are not met."11
Our intuition is that all of 7, 8-8.2 are true (these are my elaborations):
7. He says that he will kill thousands of people if his demands are
8. He says that he will detonate the bomb if his demands are not
8.1. He says says that he will create mayhem downtown if his
demands are not met.
8.2. He says that he inflict great damage on our community if we
don't do as he says.
Cappelen and Lepore (1997) summarizes these point as follows: "...indirect reports
are sensitive to innumerable non-semantic features of reported utterances and even
on the context of the report itself. As a result, typically there will be indefinitely
many correct indirect reports of any particular utterance." (Cappelen and Lepore
(1997), p. 291). Soames draws a related conclusion:
[The phenomenon of many propositions being expressed by an utterance of
a sentence] "…is an extremely general one that has nothing special to do
with proper names, indexicals or any of the semantically contentious issues
that are of special concern here. On the contrary, the phenomenon of
asserting more than the semantic content of the sentence one utters in a
context is all but ubiquitous. … what an assertive utterance of a sentence s
counts as asserting depends not only on the semantic content of s, but also
on the obvious background assumptions in the conversation and the
speaker's intention about how the speaker's remarks is to be interpreted in
the light of them." (Soames, 2002 pp. 76-78)12
SPAP can be developed in many different ways. Two issues are particularly important for
how one thinks about the semantics for "know": The various kinds of contextual
variability that affects what is said and the role of semantic content. I discuss these in
SPAP Elaboration 2: Contextual Variability of Speech Act Content
According to SPAP, three kinds of contextual variability affect speech act content:
V1 Variability in what is asserted / said by an utterance of S from one context of
utterance to another: All versions of SPAP agree that the set of propositions asserted
by an utterance of a sentence, S, can vary from one context of utterance to another.
The set of propositions said and asserted by utterances of, for example "A will
detonate the bomb if his demands are not met" will vary from one context to
V2 Variability in what is said / asserted by an utterance of a sentence from one
context of interpretation to another. A more radical version of SPAP claims that
what's said by an utterance u of a sentence S in a context of utterance C, will vary
between contexts of interpretation. On this version of SPAP, what's said by utterance,
u of S in C, relative to a context of interpretation C', might be different from what's
said by u relative to another context of interpretation C''. This is the view endorsed by
Cappelen and Lepore (1997) and considered, but not endorsed, by Soames (2002) and
Cappelen and Lepore (2005).
V3 Variability in what is saliently said / asserted from one context of interpretation to
another: Not all proponents of SPAP agree to the kind of content relativism described
in V2 above. However, all versions of SPAP would agree that what is saliently
asserted by an utterance u (in a context of utterance C) relative to one context of
interpretation CI might be different from what's saliently asserted by u from relative
to another context of interpretation, CI*. Any version of SPAP will need a theory
about how one part of the speech act content becomes salient in a context of
interpretation. According to SPAP an utterance asserts many propositions. One
(maybe several) of these will be more salient than the others to the speaker and
interpreter. Whether or not a SPAP proponent accepts V2, she will accept the view
that the saliently asserted proposition might vary from one context of interpretation to
One of the main objections to skepticism is that it can't account for intuitions we have
about what speakers say and the ways these intuitions vary between contexts of utterance.
In responding to those kinds of objections appeals to V1-V3 will prove useful (more
about that below.)
SPAP Elaboration 3: Role of Semantic Content
According to some versions of SPAP, any utterance of a sentence, S, says/asserts
the semantic content of S (even though that proposition might not be saliently asserted
relative to all contexts of interpretation.) According to other versions of SPAP this is not
the case -- there are utterances of S that do not assert the semantic content of S. In what
follows, I'll assume that the semantic content is always asserted.
Summary: SPAP and "know"
Applied to sentences of the form "S knows that p" these three elaborations of SPAP
have the following implications: Let u be an utterance of a sentence of the form "S knows
that p" in a context of utterance C. According to PS:
a) One proposition said / asserted by u is the proposition semantically expressed
(call this proposition p). p is false (that's the implication of endorsing skepticism).
b) u might assert many propositions, p1…pn in addition to p.
c) If you endorse V2 above, p1…pn will vary with the context of interpretation: i.e.
interpreted from context of interpretation CI the set of propositions asserted by u
might be different from the set of propositions asserted relative to another context
of interpretation, CI*.
d) Even for those who don't endorse V2 above, it follows from V3 that the salient
component of the speech act content will vary from one context of interpretation
to another, i.e. which one of p and p1…pn is salient will depend on the context u is
I've gone into this much detail about the SPAP component of PS because it is an
unfamiliar view and because it is at the center of the defense of skepticism presented
below. In what follows, I first respond to the charge that skepticism implies some kind of
error theory about intuitions about the semantic content of our views. In the last part of
the paper I respond to the charge that skepticism is incompatible with the knowledge
account of assertion.
2. Pluralistic Skepticism, Linguistic Intuitions, and Contextual Variability
Our linguistic behavior, our intuitions about what speakers say and our intuitions
about the truth-value of what they say seem to be sensitive to contextually variable
features in a way that might seem difficult for a skeptic to explain. Derose says:
"In some contexts, “S knows that P” requires that S have a true belief that P and
also be in a very strong epistemic position with respect to P, while in other
contexts, the same sentence may require for its truth, in addition to S’s having a
true belief that P, only that S meet some lower epistemic standards" (Derose
What makes for this difference? In the examples favored by the contextualist it is various
practical factors (such as what is practically at stake) that vary between contexts of
utterance. In other words, this sensitivity to contextual standards is not just brought out
when thinking about skeptical possibilities -- it is not just when speakers are in
philosophical contexts that their standards shift. As Derose points out:
"To make the relevant intuitions as strong as possible, the contextualist will
choose a “high standards” case that is not as ethereal as a typical philosophical
discussion of radical skepticism … it makes the relevant intuitions more stable if
the introduction of the more moderate skeptical hypothesis and the resulting raise
in epistemic standards are tied to a very practical concern, and thus seem
reasonable given the situation." (Derose (2002), p. 191.)
There are two kinds of variability that the skeptic is asked to account for:
Let S be the sentence "A knows that p (at t)", and let u and u' be two utterance of
S in C1 and C2, respectively. Let C1 be a so-called low standard context and C2 a
high standard context. According to the contextualist our intuitions about the
truth-values of u and u' might differ. We can, for example, have the intuition that
u is true while u' is false. This can't be explained by a change in A's epistemic
position (both utterances are about A at time t). The relevant difference, according
to the contextualist, is can be found in the attributors' practical situation. The only
relevant difference between the contexts might be, for example, that the speakers
have different practical concerns, no philosophizing or strange skeptical
possibilities need enter into the story.
Our intuitions about what is said by utterances u and u' of S can varies (let S, u
and u' be as in (i) above). Our intuition might tell us that u' says that S has a true
belief that p and is in a very strong epistemic position with respect to p, while u
says that S has a true belief that p and is in a less strong epistemic position with
respect to p14. Again, the only difference between the contexts is that the speakers
have different practical concerns.
The problem for the skeptic is supposed to be this: If the semantic content of "know"
invokes a super-high standard and this is invariant between contexts of utterance, the
semantics will provide no explanation of the variability in (i) and (ii). And, it is assumed,
these are the kinds of intuitions that a semantic theory for English should account for.
Reply: How PS Deals with Contextual Variability
Not only can PS easily explain this kind of data, but it can do so better than any
alternative theory. The assumption has been that the skeptic has to say that these
intuitions are, somehow, mistaken, i.e. has to defend some kind of large-scale error
theory about speakers' intuitions. A skeptic who endorses PS, however, is not committed
to an error theory. According to PS, we assert many different propositions when we utter
sentences of the form "A knows that p". We do not just assert the proposition
semantically expressed. So a proponent of PS will hold that some of the propositions
asserted (said, claimed, etc) by an utterance of a positive knowledge attribution can be
true even though the proposition semantically expressed is false. She will also hold that
the totality of asserted propositions can vary from one context of utterance to another,
and from one context of interpretation to another. So there are plenty of resources in PS
to account for the kind of variability appealed to by contextualists. PS, by virtue of
incorporating SPAP, predicts exactly this kind of variability.
Before going into more detail about how PS accounts for (i) and (ii), one
important methodological remark: If SPAP is correct, it's a dangerous simplification to
talk simply about the intuitions we have about what's said by an utterance, u in a context
C (or about our intuitions about the truth-value of what was said by an utterance u in a
context C.) All such intuitions must be relativized to a context of interpretation. We
should always talk about our intuitions about u from a context of interpretation C. Not
doing so gives the impression there's some neutral and (possibly) privileged semanticist's
point of view from which we can have intuitions about utterances. If SPAP is correct,
there is no such privileged standpoint.
To see how PS deals with these kinds of cases, I distinguish (as the contextualist
does) between three kinds of epistemic standards that can be invoked: low-standards
(LS), ordinary-high standards (OHS) (not the super-high philosophical standards, but the
kinds of standards that, according to contextualists, are invoked by raising of ordinary
non-philosophical / practical stakes), and Super-High standards (SHS) (the kinds of
standards triggered by skeptical arguments.)
The challenge is to explain why our intuitions vary between contexts in the way
described by (i) and (ii). Here are some schematic explanations, using the resources of PS
-- in all the cases below let S be the sentence "K knows that p (at time t)".