Declaratives Are Not Enough
Author(s): Nuel Belnap
Source: Philosophical Studies: An International Journal for Philosophy in the Analytic
Tradition, Vol. 59, No. 1 (May, 1990), pp. 1-30
Published by: Springer
Stable URL: http://www.jstor.org/stable/4320114
Accessed: 28/05/2009 14:42
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=springer.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].
Springer is collaborating with JSTOR to digitize, preserve and extend access to Philosophical Studies: An
International Journal for Philosophy in the Analytic Tradition.
http://www.jstor.org
NUEL BELNAP
DECLARATIVES
ARE NOT ENOUGH
(Received17 January,1989)
1. THE DECLARATIVE FALLACY
My thesis is simple:systematictheoristsshouldnot only stop neglecting
interrogativesand imperatives,but should begin to give them equal
weight with declaratives.A study of the grammar,semantics, and
pragmaticsof all three types of sentence is needed for every single
serious programin philosophythat involves givingimportantattention
to language.1
Part of the backgroundof my thesis is that in our culture when a
logician, or nearly any trained philosopher, says 'sentence,'what is
meant is a declarativesentence,2a sentence capable of having,as they
say, a truth-value,or maybe truth-conditions,a sentence that can be
used to 'say'something,a sentence expressinga proposition,a sentence
that can play a role in inferenceas eitherpremissor conclusion,a sentence that mightoccur in someone's (say Quine's)'canonicallanguage.'
This is what is to be rejected.This is the DeclarativeFallacy.Instead,
one should recognize that from the beginning there are not only
declarativesentences,but, at least, both interrogativesand imperatives.
The grammariansare right and those teachersof elementarylogic that
seem to have miseducatedmost of us are wrong:give all sentences
equal time, and do not take declarativesas a paradigmof what can
happenbetweenfullstops.
I wish eagerly(but parenthetically)to grantthat there are or may be
other sorts of sentences besides the declaratives,interrogatives,and
imperatives,say the optatives,or the performatives,and indeed further
on I will ask you to think a moment about the precatives,but it is no
part of my thesis that I've got the goods on what sorts of sentence are
enough;so if you just promise to take my remarksas nonexclusive,we
canmakesome honestheadwayandall willbe well.
PhilosophicalStudies59:1-30,1990.
? 1990 KluwerAcademicPublishers.Printedin the Netherlands.
2
NUEL BELNAP
1.1. DeclarativesAre Not Enough
Here is an example of the DeclarativeFallacy. Frege says, rightlyor
wrongly,that only in the context of a sentence words have meaningthe famous context principle.3There is, I think,small doubt that Frege
himself,and no doubt that the traditionthat followed him, has in mind
only declarativesentences,4leaving out the interrogativesand imperatives altogether,and if so, then the context principleis bad philosophy.
For one thing,to the extent that it is true it is seriouslymisleading,for
the role that words play in interrogativesand imperativesis at least as
importantas the role they play in declaratives.Thus, the word six can
obviouslybe just as meaningfulin an interrogativeor in an imperative
as it is in a declarative.And conversely, if you want a contextual
explanationof the meaning of six, the declarativecontexts are not
enough:you had better know as well how it functionsin interrogatives
andimperatives.Declarativecontexthas no prideof place.
You may respondthat once you know all about six as it functionsin
declaratives,then what it comes to in the context of interrogativesand
imperativesis determinedand thereforesecondary.The point is doubly
wrong.In the first place, it is a cheap philosophicalshot, for symmetrically, if I know all there is to know about six as it functionsin imperatives, or in interrogatives,then it is to an equal extent determinedwhat
six comes to in declaratives.If for instance I know everythingthat
anyone can ask using six, then I know everythingthat anyone can say
using six. And in the second place, some words,and six is one of them,
playdistinctiverolesin interrogatives,as in
Whichsix speechactsaremostimportant?
Here the six is arguablypart of the interrogativeform ratherthan part
of a declarativematrixsuggestingpossibleanswers,as it mightbe in
Whichspeechactsrequirethepresenceof sixpersons?
For a second thing, there are certainwords or modes of combination
whose significanceis principallyto be gathered from their roles in
interrogatives,say the questionwordsthemselves.Take whatas in
Whatis an illocutionary
force?
DECLARATIVES
ARE NOT ENOUGH
3
Frege'scontextprinciplesuggeststhat what has no meaningexcept as it
occurs in a sentence, but the primarysort of sentence in which what
occurs is not a declarativeat all. If you wantto know what what means,
look at it in the context of interrogatives- that is the best of advice.
Another example is offered by the way or ways or functions in
how or worksin, for example,
interrogatives;
Is it declarativesor interrogativesthat have invertedword
order?
does not have much to do with truthtables or assent tablesor anything
like those devices for understandingdeclaratives,because it has to do
with the interrogativeform itself. I am supposingthat it is obvious to
you thatwe arenot givena yes-no questioninvolvingthe declarative,
haveinvertedwordorder.
Declarativesor interrogatives
to deterInstead the or in the interrogativeis working interrogatively
minewhatis to countas a possibleanswer.
Let me summarize:declarativesare not enough. They are too
insubstantialto count as a paradigm for interesting theses in the
philosophyof language.
1.2. PropositionsAre Not Enough
Here is an easy corollaryof or additionto my thesis:assertions,that is,
the speech acts so called, are not enough.And it needs bearingin mind
that not only are they not enough,but they are not in any philosophical
sense 'primary,'even though the canonicallanguagesof the dominant,
more formallogiciansallow for nothingelse. To suppose that assertion
is the primaryspeech act is to commit at least a misdemeanorwith
respectto the DeclarativeFallacy.
Well, you may say, avoidance of the Declarative Fallacy is tired
advice that you have takenall your life, becauseyou are a fan of speech
acts, and all speech activistshave alwaysknown that there are numerous kinds of speech acts, with non-assertionssuch as questions and
commands being prominent among them. I, too, am awed by our
magicalpowers to do things with words, but alas, many of the central
4
NUEL BELNAP
theoristsin this traditionare just as guiltyof the DeclarativeFallacyas
those philosophers, say, who exclude interrogativesand imperatives
fromtheirinvented'canonicallanguages.'
For example,permitme to quotefrompage 1 of a recentbook.5
The minimalunits of human communicationare speech acts of a type called illocutionary acts. Some examples of these are statements, questions, commands, promises,
and apologies. Whenevera speaker utters a sentence in an appropriatecontext with
certain intentions,he performs one or more illocutionaryacts. In general an illocutionaryact consistsof an illocutionaryforceF anda propositionalcontentP.
On the surfacethis looks to be a paradigmrejectionof the Declarative
Fallacy,but it is not. The fallacyis, at the very beginningof the theory,
to endow every kind of speech act with the same kind of content,here
called 'propositionalcontent.'Let us say that each assertion,or whatwe
might call 'declarativeact' (this would not be Searle-Vanderveken
terminology,but it is just as good) consists of an illocutionaryforce F
and a propositionalcontent P. Then you are guilty of the Declarative
Fallacyif you suppose that interrogativeacts, or imperativeacts (as we
might call them), can have the same propositionalcontent, P, as a
declarativeact. Of course the whole drivingidea of speech activistsis
that a single content can be clothed in a varietyof differentforces, and
it is not this crucial idea about which I am now complaining.I am
objecting only to the DeclarativeFallacy, which here emerges as the
special case of supposing that interrogativeacts and imperativeacts
have the same content as declarativeacts. I understandthat there are
many speech acts that share a propositional content but differ in
illocutionaryforce; good. And there are many other speech acts that
share an interrogativecontent but differ in illocutionaryforce, and
others that sharean imperativecontentbut differin illocutionaryforce.
So the programis a healthyone; the only - but serious - mistakeis to
suppose that you can identify the content of all speech acts with
propositionalcontent, that is, with the content of declarativespeech
acts or assertions.
One way to avoid this mistakeis to take the content of a speech act
to consist not of a propositionall by itself, but instead to consist of a
proposition togetherwith a marker of mood or perhaps force - to
count the force, so to speak, as part of the content.6But after all this
strategy re-commits the Declarative Fallacy, for even with its mood
DECLARATIVES
ARE NOT ENOUGH
5
marker,at bottom each speech act is construedas based on a proposition, situated,so to speak, at the core of its core. Strictavoidanceof the
DeclarativeFallacy, however, requires the recognitionthat interrogatives and imperativesare not just markeddifferentlyfrom declaratives,
butpossessfundamentallydifferentunderlyingcontentstructures.
In a word,propositionsarenot enough.
I have to make good on an implied promise to distinguishthe content of interrogativeacts and imperativeacts from that of declarative
acts,butnot now.7
Again, summarizingso far: declaratives,assertions,and propositions
are not enough. They do not provide enough variety of content to
supportthe content/forcedistinctionof speechact theory.
1.3. TruthConditionsAre Not Enough
There are yet more who thriveby committingthe DeclarativeFallacy.
There is an enormous school with an even more enormousgroup of
hangers-onthat says that what we ought to pay attentionto are truth
conditions,and that to have a theoryof truthconditionsis to have it all
- well, most of them say, grudgingly,that perhapswe also need a small
auxiliarytheory of reference,just in case the languageis non-canonical
enough to contain a few singularterms. The reason this school is so
large is not entirelya matter of charisma;in fact there is tremendous
enlightenmentto be had by thinkingof various pieces of languageas
resultingby grammaticalcombination,and seeing how the meaningof
the grammaticalwholes arise out of the meaningsof their parts.Tarski
was a genius. Nevertheless,the slogan that meaningis truthconditions
is flawedin more thanone way, not least becauseit seems to force us to
take the concept of truthmuch more seriouslythanwe should,nor least
because it appearsto suggestthat our understandingof languageresults
from internalizingthe recipes that Tarski invented for the first-order
functionalcalculus.
But I am puttingthese flaws aside in order to call to your attention
that the only items that can possibly have truthconditionsare declarative sentences,or at least items with a propositionalcontent.Davidson's
famous and deceptively short transition from meaning to truth8 is
remotelyplausible only for the meaningof declarativesor their ilk. If,
6
NUEL BELNAP
as arising
however,we wantto understandthe meaningof interrogatives
by compositionfrom the meaningsof theirgrammaticalconstituents,as
we should, then truthconditionsare not enough.What we want when
we want to become clear on an interrogativeis what question it asks,
and what counts as a possible answerto it; and in the frameworkthat
approachesthis problemin the spiritof Tarski,we wantan accountthat
sees this dimensionof meaningas arisingby grammaticalcombination.
Furthermore,and with equal importance,just as with Tarskiwe investigatethe ways that declarativesentencesare grammaticalparts as well
as wholes, so that they contributetheir meaningto largercontexts,just
so we should expect the same of interrogatives;we should expect them,
in an adequatelanguage,to contributetheirmeaningto otherinterrogatives, or to declaratives,or to imperatives,or whatever.We should ask
for a Tarskiaccountof the declarative,
How importanttruth conditional semantics is said to be
dependson whomyou ask
that makes it clear that in that declarativethere are embedded two
ingredientinterrogativesentences, each of which should contributeits
distinctivemeaning.Let me emphasizethe point:interrogativesoccur as
compositional elements in declarativesjust as truly as declaratives
occur as compositionalelements in interrogatives.Therefore,whether
you believe that understandingclimbs the grammaticaltree compositionally, or descends the grammaticaltree contextually,you should
agree that interrogativesand declarativescannot have independent
theories.
There is of course more to interrogatives,indeed a great deal more.
Here is a tiny samplethat arose in the work that some of us, including
Bennett9 and Thomason, did together a number of years ago: in
thinkingabout
Michael wonderedwhere each adequate theory of imperativesispublished,
we shouldsee the embeddedandthereforecontributoryinterrogative,
whereeach adequatetheoryof imperativesispublished,
as itselfarisingby a universal
DECLARATIVES
ARE NOT ENOUGH
7
each adequatetheoryof imperatives
quantificationinto the open interrogative,
wherex ispublished.
That is, we have here a quantifier expression transformingnot a
declarativeinto a declarative,but an interrogativeinto an interrogative.
For understandingthese transformations,
truthconditionsarenot apt.
It is also obvious that the nature of the complaintsI am making
counts againstthe sufficiencyof 'verificationconditions,'such as those
urgedby Dummett,'0just as much as they do againstthe sufficiencyof
truthconditions.It is the very typeof the conditionsthatis wrong.
The same holds for imperatives,though to a lesser extent, for the
content of an imperative act is certainly more closely allied to a
propositionthan is the content of an interrogative,and thereby more
mold. Also
easily fits the truth-conditionalor verification-conditional
much less is known about the compositionalsemanticsof imperatives.
In the first place, not much is knownof how the meaningof an imperative arises out of the meaning of its constituents,although various
essays in action theorycan perhapsbe takento be contributionsto this
theory. In the second place, philosophers have almost universally
ignored the obvious fact that imperativesembed in largercontexts as
readily as do declaratives,so that no one at all has studied how the
meaning of an imperative contributes to those larger combinations.
Whatis it exactlythatyou haveto knowaboutan imperativesuchas
John,giveus a lectureon truthconditions
in order to say something interesting about its contributionwhen
embedded,as in the following:
Mary,requestJohnto giveus a lectureon truthconditions?
A signalweaknessof the speech act program- not in its essence but in
its present state - is its failure,in spite of its attentionto imperatives
andinterrogatives,to focus on the problemsof compositionality.
Summary so far of what is not enough: declaratives;assertions;
propositions;truthconditions;verificationconditions.None of them are
enoughfor a compositionaltheory of meaning.In part the thesis is that
understandinghow such a theory works - or doesn't work - for
8
NUEL BELNAP
declarativesis not sufficientfor understandingthe deepest featuresof
compositionality.It is past time for the dialectic about compositional
theoriesto move to a new andmorephilosophicallyadequateplane.
1.4. InferenceIs Not Enough
And there are the inferentialists.Instead of attemptinganalyticallyto
see meaning as arising by composition, an inferentialistgives an
explanationof the meaningof a declarativeform in termsof its role in a
larger context. Nor for the inferentialistwill satisfactionbe found by
confining attention to just the context of a single illocutionaryact,
definedby Searleand Vanderveken,as you recall,as a 'minimalunit of
humancommunication'(emphasissupplied).Insteadthe appealis to the
larger context of inference,where the declarativecan figure either as
conclusionor as premiss.Insteadof seeing a declarativeas inheritingits
meaninganalyticallyfrom its parts,one sees a declarativeas derivingits
meaningcontextuallyby the role it plays in inference.I have to tell you
that this now-popularapproach,to the extent that it claims to tell us
philosophicallythat we know our way aroundlanguage,is also miserably guiltyof the DeclarativeFallacy,for, to a first approximation,it is
only declarativesthat can figurein inference,and we are therebygiven
no purchaseon interrogativesor imperatives.
My case is strongestfor premisses:for decades logiciansled astray
by the Declarative Fallacy have tried out little, tiny examples of
inferencesinvolvinginterrogatives,suchas
Who has a good theory of interrogativesand who wants
one?, therefore,Whohas a good theoryof interrogatives?
But never, ever has anyone even suggested a long or interesting
inferentialchain of interrogatives,either in an idealizedlanguageor in
plain English. Nor has anyone ever suggested that we could learn
somethingabout the meaningof an interrogativeby its use in such a (I
shudderwith quotes) 'inference.'Nor is it surprising;since the meaning
of an interrogativeis not essentiallypropositionalor truth-conditional,
it is hardlylikelyto be inferential.
Analogously, using imperatives as premisses can hardly be con-
DECLARATIVES
ARE NOT ENOUGH
9
sidered idiomatic.Maybe we sometimessay with the deontic logicians
fromA. Ross "Ito B. Chellas,12
Mail the manuscript;so mail it or burnit.
But I hope not very often. Certainlynot often enough to enable us to
work out what an imperativemeans by examiningits role as a premiss,
or, for those of us who thinkthat Gentzenwas on to something,to find
a cut-eliminationtheoremfor imperatives.
As a possible objection,accordingto a long-standingtraditionin the
theory of action, agents plan what they do by resortingto practical
syllogismsor enthymemeslike
I shall burnthe manuscript;therefore,I shallset it onfire.
While much of the talk about practicalreasoningtends to confuse me,
this episode of practicalreasoningsurelyinvolves stand-alonedeclarative expressions of intentions rather than stand-alone first-person
imperatives,which is not to say that declarativesthat carryexpressions
of intentionsdo not containembeddedimperatives.
With respect to conclusions,the situationis not obviouslysymmetric
betweeninterrogativesand imperatives.On the one hand,it is perfectly
straightforwardto place an imperativeas the conclusion of an inference, sayin givingadvice.The consecution
Truthconditionscausecancer;therefore,avoidthem
is just fine. Interrogatives,on the other hand, when they appear as
conclusions, seem never to contributeonly their content but also in
additiontheirstand-aloneforce.
Considerthe perfectlyidiomatic
Quine does not avoid the declarativefallacy; therefore,who
does?
and contrastit with the precedingexample,in whichthe premissgives a
reason for the content of the conclusion, a reason to avoid truth
conditions.But the premissin the quine exampledoes not give a reason
for the contentof the conclusion,who avoids the declarativefallacy,but
only a reason for asking that question.'3To see this point more clearly,
10
NUEL BELNAP
notice that the ostensible interrogativeconclusion can be paraphrased
by the followingimperative:
Quine does not avoid the declarativefallacy, so tell me who
does.
Furthermore,even though imperatives can occur as conclusions in
inferentialtrainsof justification,it is not a good strategyto pin all their
meaning to such a role. The reason for this derives from a crucial
observationof Hamblin:`4some among all imperativeacts are acts of
advice or warning,say the advice or warningnot to ignore interrogativescarriedby
Don'tignoreinterrogatives.
Such imperativescall for justification.But other imperativeacts are
what Hamblincalls 'willful,'being either for example the issuing of a
rule againstignoringinterrogatives,or perhapsa plea or a requestnot
to ignore them, also issued with exactly the same imperativesentence,
Don't ignoreinterrogatives.
Here, thoughtheremay be a call for reasons
for the act of issuingthe imperative,there is no call for reasonsfor the
content of the imperativeitself. And yet the content of the different
imperativeacts of advisingor warningor pleadingor requestingis quite
the same, whetherwillfulor not, a samenessthat our theoryof imperatives should recognize by not looking to inference to confer all their
meaning.It is easy to become confusedabout this if we are too swift to
label imperatives'exit moves' in the languagegame;my advice is: don't
becomeconfusedin thisway.
Summaryof what is not enough: declaratives,assertions,propositions, truth or verificationconditions, and inference.No philosopher
should be deluded into thinkingthat he or she knows the way around
the declarativesof our languagewithout understandingtheir dependence on interrogativesand imperatives.And no philosopher should
delude others into supposingthat a method such as that of compositional semantics,or that of conceptualor interpersonalcontextualroledescriptions,can arguablybe defended as philosophicallyadequateor attackedas inadequate- even as a program,unless it is shownhow
it works - or doesn't work - for interrogatives and imperatives.
So, you might ask, what is enough?In my opening remarksI asked
DECLARATIVES ARE NOT ENOUGH
11
you not to ask me that.Instead,permitme to use this list of what is not
enough as an indication of what more is needed. Needed for what?
Needed to make us see that nowhere in that part of philosophy that
relies on theories about language can what we know so far about
declarativesreasonablybe taken as a comfortingparadigmgivingus to
believe that 'the rest is somethinglike that';so that if only we understand declaratives,we can be sure we know our way aroundour mode
of being as it is exhibitedin language.Interrogativesand imperatives
are not 'somethinglike declaratives.'What they share with declaratives
is this:that unless we bringthem to the light,or the light to them, there
is much else of philosophicalimportancethat will also remainin the
dark.
My plan for the following sections is to say a few things about
grammar,semanticsand pragmaticsof interrogativesand imperatives,
hoping by their inadequacyto push you to include these neglected
forms in all your orisons, and never again carelesslyto say 'sentence'
when more narrowly you mean 'declarative,'or 'proposition'when
more widely you mean 'contentof a speech act.'And so to move down
closerto the tacksthatarebrass.
2. INTERROGATIVES MAY NOT BE ENOUGH, BUT THEY HELP
I have had my say about interrogativesin variousplaces, and here wish
only to fill out the currentperspectivewithsome briefremarks.'5
2.1. GrammarforInterrogatives
Independentlyof whateverkind of grammarwe endorse for declaratives, we should articulatedifferent grammaticalstructuresfor interrogatives.A few logiciansin the last thirtyyears or so have worked a
little at laying out normativelywhat a good grammarof interrogatives
should be, and a few good linguists have worked at the descriptive
grammaticaltheory of interrogatives,whereby it is crucial that by
'interrogatives'I signify not only stand-aloneinterrogativescapable of
carryingspeech acts, but also constituentinterrogatives,16
sometimes
called 'indirectquestions,'capable of being embedded in larger contexts. The chief point to stress is that the grammarof interrogatives
12
NUEL BELNAP
ought to resist oversimplificationby tired philosophersout to make a
quickreduction,or a career.Let a singlebriefexamplesuffice:
Whowasthe authorof How to do thingswithwords?
is, for all its apparentsimplicity,grammaticallyambiguous,and that in
at least two ways.In the firstplace, the scope of the definitedescription
operator, the, could be either wide or narrow,just as Russell would
have said; but this sort of thing is familiarand I wish to leave it. The
other grammaticalambiguityis this: Whowas the authorof How to do
thingswith words? can derive by the introductionof the questionword
who into eitherof two quite differentdeclarativematrices.Considerthe
following,
x wasthe authorof How to do thingswithwords,
and
the authorof How to do thingswithwords wasF,
in which I intend that x occupy the place of a nominal singularterm,
and that F occupy the place of a predicate adjective or a predicate
nominal.That is, who may be thoughtto be proper-name-like,or termlike, or adjective-like,since all of these may be sensiblycombinedwith
is. We should not assume that these differentderivationsgive rise to
exactly the same question. On the first grammaticalderivation,the
interrogativeinvitesas an answer,
Russellwasthe authorof How to do thingswithwords,
while on the second grammaticalderivation,what is invited is somethinglike
The authorof How to do thingswith words was a fat, early
fourteenthcenturyFrenchauthorfromProvence.
On the first derivationthe question asked amounts(for a logician)to
somethinglike
Whichperson - and please confine yourself to canonical
and very rigid designators- is identical to the author of
How to do thingswithwords?
DECLARATIVES
ARE NOT ENOUGH
13
On the secondderivation,the questioncomes to
Whatare a few interestingpropertiesof the authorof How to
do thingswithwords?
It is the fact that interrogativesinvert word-order that causes this
particulargrammatical- I stress that it is a matter of grammarambiguity.Whetheror not you agree that I have correctlyrepresented
the differencebetween the two interrogatives,you will certainlyagree
thatwhatI sayis basedon grammar.
This point is both contentiousand tiny, and it was meantto be both,
so thatit could become more palpablewithoutthe need of a substantial
apparatuswith which firmlyto grasp it; but I hope the point is at least
large and brightenough to suggestthe error of thinkingthat we clearly
understandwhat is meant by an arbitrarywho-question(much less a
what-questionor a why-question),and the error of supposingthat we
have no need for a grammaticaltheory of a philosophicallysuggestive
declarativethatcontainsan embeddedinterrogativesuchas
George IV did not know who the author of How to do
thingswithwords was.
We should all be saddened,incidentally,by how many of us have been
trained to think about the meaning of that declarativeor one of its
cousins without being brought to notice that it contains a constituent
interrogative.
So much for the claim that philosophy needs a grammarof interrogatives as much as it needs its endlessly elaborated grammarof
declaratives.'7
2.2. Semanticsfor Interrogatives
Interrogativesdeserve a compositional semantics that is not piggybacked on the semanticalcorrelatesof declaratives.But if truthconditions won't do, what else should a semantic theory for interrogatives
draw on? The answergoes back at least to Hamblin,18and more than
once I myself have helped to spread his word: instead of truth conditions, interrogativesneed answerhoodconditions.If you are persuaded
that there is enlightenmentto be had about a declarativeby learning
14
NUEL BELNAP
how its truth conditions arise out of the meaningsof its constituents
and its structure,and if you are persuadedthatit is the truthconditions
of declarativesthat are needed when it comes to embeddingthem in yet
largerstructures,then with perfect analogy,you should expect to find
correspondingenlightenmentin seeing how what counts as an answer
to an interrogativearises out of the meaningsof its constituentsand its
structure,and you should correspondinglyexpect that it is the answerhood conditionsof an interrogativethat are needed when it comes to
embeddingthemin yet largerstructures.
Let me illustratewith a small problematic.What if anythingis the
difference between the meaning of the following two interrogatives,
whetherstand-aloneor embedded?
Whichlogicianusesdeclaratives
for trainingherdog?
Whichperson who uses declarativesfor trainingher dog is a
logician?
I tell you by the method of authoritythat people disagreeas to whether
or not these two interrogativesdifferin meaning,and I am not going to
try to convinceyou one way or another.Rather,I simplysubmitthat if
we are to find a differencein meaningbetween these two, we should
forget about truth conditions or verification conditions. The best
strategyis to look to see if they have differentanswerhoodconditions;
thatis all I wishto urge.
If we like to reify the resultsof semanticinquiry,as indeed I thinkis
always helpful in keeping us clear and even honest, then we need a
contentfor interrogatives,a content that is distinctfrom theirforce. As
a word I like 'question'for the contentof an interrogative,but for those
disinclined towards any new words, I am happy for now with just
'interrogativecontent' for the content of an interrogative.I see it
possible to have as many theories about interrogativecontent as there
are theories about content of answers, includingjust as many nominalistic theories or non-theories.Certainlymany and many, including
manyfalse or unhelpfulones, have been triedout over the decades;but,
and this is the reason for the being of my present remarks,the vast
philosophicmajorityjust ignoresthe matteraltogether,and that is what
is really bad philosophy. My own view is that we should take the
content of an interrogative to be
-
put circularly
-
the property of
DECLARATIVES
ARE NOT ENOUGH
15
being an answer to the interrogative.What will count as an answer
varies from context to context, so that, for instance,what counts as an
answer to the question of which logician uses imperativesfor training
her dog depends on who, in a given context,qualifiesas a logician,but
does not depend on who in that context uses declarativesfor training
her dog. Onlythe trueanswerdependson that.
2.3. Pragmaticsof Interrogatives
Inference,as I pointed out above, does not suffice to characterizethe
meaning of interrogatives.But what can possibly be enough for
interrogativesif inferenceis not? We have to stop to introduceat this
point a subtlety depending on force versus content, a subtlety that
alreadyarises for declaratives.We know throughtrial by Gentzen that
if all that we are interested in is content as it is passed through
embedding,then a ratherabstractnotion of inferencewill do; we need
little more thanthe divisioninto premissesand conclusion,and thatit is
meaningfulto separate good inferences from bad. But if we want
enlightenmentabout force, say assertive force, and especially if our
philosophicalinclinationis pragmaticallyto derive content from force,
then we shall need to follow Brandom'sSellars-inspiredlead in seeing
inferenceas more than a mere abstractsemanticrelation,and as rather
a part of a normativestructureinvolvingat least the undertakingand
The analogyis that for a purelysemantic
attributionof commitments."9
theory of interrogatives,hardly more is needed than the abstract
concept of the answerhoodrelationin place of the inference relation;
but that for a decent understandingof the force of interrogatives,
whetheror not we can thereby derive an account of their content, we
shallneed an appropriatenormativestructure.
I am describing seldom trod territory, and certainly territory in
which I am myself not at home, but territoryworth exploring. Of
course, when we put a question with an interrogativewe commit ourselves to its presupposition,if any;that'seasy. And surelywe put on the
conversationaltable, whateverthat means, the-possible answersto the
questionwe ask. What else we do seems to be variousin the extreme;
let me observe just for example that an interrogativeact can be either
an injunction- occasionallyeven a 'command'if you like, as when the
16
NUEL BELNAP
DistrictAttorneyquizzes a sworn witness - or a request,as when you
ask the time of a passer-by;and in fact it may be that the varietiesof
interrogativeact are not fewer than the varietiesof imperative,which
are legion. Nevertheless,I cannotleave this topic withoutone last piece
of the manifesto:it seems to me extremely unlikely that one can
develop a philosophicallyadequate interpersonalnormativestructure
for assertionswithoutsimultaneouslytreatingof questionsin the sense
of interrogativeacts.
The inferentialist,and Hamblinexplicitlyin his theory of dialogs,2"
pictureseach person as carryinga slate on which is inscribeda list of
declarativesrepresentingthose propositions to which the person is
committed.But propositionalcommitment,either undertakenor attributed, is not enoughexhaustivelyto characterizethe contentof the slate.
For one thing,there is just too much non-trivialtruthin R. G. Collingwood's slogan that every propositionis an answer to a question.2'At
some point in describingthe game of conceptualthoughtyou are going
to have to make room on the slate for a differentkind of statement
representinga differentkind of content, not declarativesrepresenting
propositions at all, but interrogatives representing the questions in
which that person is privatelyinterested,or the questionsthat person is
asking,or has been asked. There seems no conceivablepossibilityof a
conversation or even a private train of thought unless there is an
expressionof the limits of what can be said next, not inferentiallybut
with regardto the categoryof importance.Whatis Annie interestedin?
Well, in who will lectureon truthconditions.That questionis what she
cares about, the answersto that question is what she wishes to gather
evidence about, and that questionis indeed what she is thinkingabout.
She will not assert anythingever, nor profit from the assertions of
others, withoutat least the traces of such interestsas can be expressed
by interrogatives(the point is due to Harrah).22Interestin questions,
like commitmentto propositions,will need to be declaredfor oneself
and attributedto others and investigatedinterpersonally.One cannot
make sense of a paradigmof canonicallinguistictransactionwithout,I
am saying,keeping track of which questions are at issue for whom. It
may even be true that whatcounts as evidencefor an assertion,or what
counts as a 'rigiddesignator,'or whethera term is being used attributively or referentially(as they say) should canonicallydepend on what
DECLARATIVES
ARE NOT ENOUGH
17
questionsare there to be addressed.If so, then inferenceis not enough
even for assertions!I hope it is but a harmlessexaggerationto assert
that there can be no mental motion, not even inference, without the
existenceof some question- not a propositionbut a question- as its
finalcause.
3. IMPERATIVES MAY NOT BE ENOUGH, BUT THEY HELP
There is so incredibly much to say about imperatives,and I have
learnedwhat little I know so recently,that it is hard to know where to
make a beginning.23
It is certainlymy view that our philosophicalcommunity is scandalouslynaive about imperatives,that our community
wallows in non-Socratic ignorance of imperatives, thinking that it
knows when it does not know, and that it mattersfor its own best
purposes whether it knows or not. It is, that is, all unknowingly
enmeshedin the DeclarativeFallacy.
WhatI have to say about imperativescomes againunderthe familiar
heads, grammar,semantics,and pragmatics;with regardto the latter I
will keep in mind both the speech activistsand the linguisticgamesmen,
those that before I called 'inferentialists'because, indulging in the
DeclarativeFallacy,they proceeded as if inferencewere the only game
in townthatis presupposedby everygame.
3.1. Grammarof Imperatives
The grammarof imperatives,especiallythe logical grammarof imperatives, is in its infancy,and whateverI say today will be gone tomorrow;
but of course that is not going to stop me for a moment. Let me
proceedby enumerating.
1. Imperativescan be either stand-aloneor embedded.In English,
just as for declarativesor interrogatives,this is a complicated
grammaticalmatter. The deepest comment is this: embedded
imperatives are in truth embedded imperatives,that is, constituent or embeddable forms of the very same imperative
sentences.In a logicallyperspicuouslanguage,they would be the
very same sign designs. Considerthe followingexamplesof the
18
NUEL
BELNAP
stand-alone imperatives 'Jack, explain sincerity conditions to
Alfred':
Mary,askJackto explainsincerityconditionsto Alfred.
MaryorderedJackto explainsincerityconditionsto Alfred.
Jack carriedout Mary'sorderto explainsincerityconditions;
or at leastJackexplainedsincerityconditionsto Alfred.
Did Mary advise Jack to explain sincerity conditions to
Alfred?
Mary demanded that Jack explain sincerity conditions to
Alfred.
Mary demanded that sincerity conditions be explained by
Jackto Alfred.
Jackrefusedto explainsincerityconditionsto Alfred.
Jack refusedMary'srequest (order, advice) to explain sincerityconditionsto Alfred.
Jack is obligated (permitted,forbidden)to explain sincerity
conditionsto Alfred.
2. The grammaticallycrucial thing about imperatives,aside from
their embeddability, is that they display an agent. As
Castafieda24and others have urged, they have the deep grammaticalform,
a to verb.
In this respect an imperativeis unlike a declarativein general,
whichmay or may not expressan agentiveproposition,and even
when it does, may not wear its agent on its surface, as the
linguistssay. But imperativesmust show forth an agent, at least
in the sense that to be understood,and (the point is crucial)to
be used in largercontexts, the agent must be uniquelyrecoverable from the surface(for example,as the addresseeof a standaloneimperative).25
3. There are various grammaticaltests for agentive declarative
sentences,butnone thatI knowdistinguishbetween
Thefire destroyedthe manuscript
and
Jackdestroyedthe manuscript.
DECLARATIVES ARE NOT ENOUGH
19
I propose the following, which although it is 'my' proposal,
derives directly from Perloff, and of course depends on the
workof others,not least on thatof Anderson.26
Thesis1. No matterthe declarativeQ,27 the sentence
a sees to it that Q
is 'agentivein a.' It may be false, true, or nonsense, but it is
alwaysagentivein a.
Thesis2. A declarativesentence, Q, expresses a proposition
'agentivein a' iff Q is accuratelyparaphrasableas: a sees to it
thatQ.28Thatis, Q is agentivein a if
Q
[a sees to itthatQ].
Therefore,for the purposes of 'logical grammar,'it suffices to
picture all agentive declaratives,and also all imperatives,as
carriedby some suchnotationas
[a stit:Q],
4.
'stit-sentence,'as I shall
differentlyin Englishdependingon how it is used.29Here
are some examples.
where we read this piece of notation
say -
Readingsinto Englishof [a stit:Q].
As a stand-alone imperative:
a, see to it that Q!
As a stand-alonedeclarative:
a sees to it that Q
a is seeingto it that Q
a sawto itthat Q.
As an embeddedimperative:
a to see to it that Q
for a to see to it that Q
that a see to it that Q30
a 'sseeingto it that Q.
As an embeddeddeclarative:
-
20
NUEL BELNAP
a sees to it that Q
that a sees to it that Q.
Thus the outstandinglyimportant grammaticalstit-facts are these,
where[a stit:QIis anystit-sentence.
1. The first blank in a stit-sentencemust receive a term for an
agent.
2. The second blank in a stit-sentence can receive an arbitrary
declarative.
3. A stit-sentenceitself is both a declarativeand an imperative:3'
it can be embeddedwherevera declarativeor an imperativecan
be embedded. For example, with regard to the former, a stitsentencecan be embeddedundera negation.
a. The result of such an embeddingis on the face of it not
itself a stit-sentence;for the special (and, to a logician,
prominent)case of negationthe resultof embeddinglooks
at leaston the surfacelike
- [a stit:Q],
not like
la stit:Q].
b. But more deeply, for the special case of negation, the
resultof embeddingis not even any kind of agentive;that
is, by our test, to which I hope you have agreed, the
declarative
it isfalse that a sees to it that Q
is not invariablyparaphrasable(or indeed equivalentin
truthvaluewith)
a sees to it thatit isfalse that a sees to it thatQ.
c. Negationwas just an example;there is no impliedwarrant
to generalizeto otherembeddingcontexts.
4. Because stit-sentencesare imperativesas well as declaratives,
they can be embedded in those special contexts fit to receive
only imperatives.
DECLARATIVES ARE NOT ENOUGH
21
5. Many contexts of serious interestto philosopherscan take only
imperatives.By our test it follows that in logical grammarwe
can withoutloss requirethat these contextsbe filled only by stitsentences. Among such contexts, the stand-alone imperative
form itself is perhaps primary.I am claiming,therefore,that if
you understand
Be at the lectureon truthconditionsat nine
as uttered with the force of, say, advice - with, that is, one of
those forces we might want to call 'imperative'- then you are
understandingBe at the lectureon truthconditionsat nine as an
as
agentive,andhence as paraphrasable
See to it that you are at the lecture on truth conditionsat
nine.
6. Furthermore,if you understandeither
Don'texplainsincerityconditionsto Alfred,
or
Refrainfrom explainingsincerityconditionsto Alfred
as an imperative(as of course it is ordinarilyand even paradigmaticallyso taken),then you must understandit as equivalent
to
See to it that you do not explain sincerity conditions to
Alfred;
or even better,since
you explainsincerityconditionsto Alfred
is doubtless an agentive and hence paraphrasableas a stitsentence,you must understandRefrainfrom explainingsincerity
conditionsto Alfredas
See to it that it is false that you see to it that you explain
sincerityconditionsto Alfred.
7. Thus, by being careful to avoid the Declarative Fallacy, we
22
NUEL BELNAP
become convincedof the accuracyof the paraphraseof refraining from acting in terms of a negated stit-sentenceembedded
withina stit-sentence:
Q]
Refrain[astit:
[a stit: [a stit:Q]]
8. The grammaticalfact, and one that should guide us philosophically, is that obligation and permission and prohibition
alwaystake imperatives.An immediateconsequenceof this train
of grammaticalthought is some modest light on deontic logic.
Hence,in logicalgrammarwe shouldneverwrite
ObligatedQ
for arbitraryQ, but only
Obligated[astit:Q],
Permitted[astit:Q],
Forbidden[astit:QJ,
etc. It makes all the difference,and leads us to numeroussmall
insights driven by our desire to avoid the DeclarativeFallacy.
Standard deontic equivalences, for example, which wholly
dependon the declarativefallacy,mustbe refined.Take
Forbidden[a stit:Q]
Obligated -[a stit:Q].
This is surely Bad Grammarsince it embeds an arbitraryand
probably non-agentivedeclarativeinto the obligation context,
whereit cannotgo. And a possiblealternative,
Forbidden[astit:Q] - Obligated[astit: - Q],
is just false when Q is not an agentive.32What is wanted is
precisely (if you think about it in your new-foundcommitment
to avoidthe DeclarativeFallacy)
Forbidden[a stit:Q]
Obligated[a stit: - [a stit:QII*
For example,the followingare preciselyequivalent:
You are forbidden to see to it that our room is filled with
smoke;
You are obligatedto see to it that you do not see to it that
our roomisfilled withsmoke.
DECLARATIVES ARE NOT ENOUGH
23
9. I want you to see that the guidelinefor deontic logic that I am
proposing is a powerful alternativeto two differentprograms.
The first and dominantprogramjust wallowsin the Declarative
Fallacyandlogicizesabout
it is obligatorythat Q
for arbitrarydeclarativeQ. The second, for example in Casand some of von Wright'swork,34sees that what has
tanieda's33
to come afterobligationis a verb,
Obligatedto verb.
This second programdoes not commit the DeclarativeFallacy,
but it also does not offer us a logicalpoint of view from whichit
is easy and naturalto see that obligation,etc., can in fact make
at least subordinatereference to declaratives.The beauty of
relyingon the schema
Obligated[a stit:Q]
is that we simultaneouslymake it easy to see that obligation
must take an imperative,and also easy to see the importanttruth
that any declarativewhatsoevercan give rise to an imperative,
and thus indirectlygive rise to the content of an obligation,by
meansof the conceptof stit.
10. There are at least dozens of other ways to embed imperatives,
includingdozens of differentforces with which to utter a standalone imperativein a standardway. There is no such thing as
'the imperativeforce.'ChapterI of Hamblin'srich book,35which
is a marvelousantidoteto the Declaratives-are-enough
Disease,
gives a wonderfully helpful discussion and table. For brief
remarkbelow I mentiononly the following:
Order[astit:Q]
Advise[a stit:Q]
Invite[a stit:Q]
Request[a stit:Q].
An importantthingto observeis that these are themselves,when
spelled out, agentives,and hence each can be paraphrasedas
stit-sentences. But we stop here because although certainly
24
NUEL BELNAP
declarativesare not enough,on the other hand, enoughgrammar
is, for now, enough.
3.2. Semanticsof Imperatives
Imperativesdeserve a rich compositionalsemantics.Such a semantics
undoubtedlywill need to go beyond what is provided by anyone for
declaratives. When working on the semantics of imperatives, the
followingquestionsareto be kept on yourlist:
1. How does such meaningas they have depend on the meaningof
their grammaticalparts? For stit-sentences, this is a rather
definite question, and a good strategywill be to try to answer
this question first; but one cannot in advance be sure that the
stit-sentenceparaphraseis sufficientto reveal all of the complexities of imperativeswith which semanticswill need to deal.
For example,whatabout all of the data and some of the theories
about by, or about the time of a killing, or about the Davidsonian strategy of emphasizing our ontology of events and
actions?
2. Here is as importanta question as any that semantictheorists
shouldkeep before theirgroupmind:what sort of meaningmust
we attributeto imperativesin order to make sense out of the
compositionalrole they play in philosophicallyimportantcontexts?Among such contexts,especiallysalientare the forces that
pertainto stand-aloneimperatives,such as advice,order, invitation, and request, but there are other embedding contexts of
importancesuch as obedience and refusal.Again I recommend
the strategyof dealingfirstwith stit-sentences,in slim hopes that
thatwillbe enough.
One thing I am sure of: the semantic representationof an
imperativemust keep the agent as a separatepart of its structure, not to be lost, for example, amid some collection of
possible worlds or truthconditionsor conditionsof correctness.
The reason is based on reflectingwhat meaningan imperative
must feed to the contexts into which it is embedded. For
example, an obligationthat Jack explain sincerityconditions is
DECLARATIVES ARE NOT ENOUGH
25
not a simple valuingof the propositionthat sincerityconditions
be explainedby Jack;it is an obligationon him. In this sense,
even if we give them truth conditions,imperativesdo not have
the samekindof meaningas declaratives.
3. Lastly,whatis the grammarand the meaning(or philosophically
interesting range of grammarsand meanings) of the various
locutions that embed imperatives,such as advice, order, invitation,request,obedience,andrefusal?
There is a fantasticamountof data about these matters,but beyond
just data, some helpful theories about some of these matters are
possible; and we can all agree that numerous people have made
contributionsto the enterprise,for example the speech activists, the
deontic logicians, and Hamblinin his book on imperatives.To begin
with, one may hope for a theory of the truth conditions of stitsentences, since after all they are advertisedas able to play a declarative role. One may entertainsuch a hope while being appalledby the
view that truth conditionsof stit-sentencesconstituteenough semantic
informationabout them in order to explainhow they embed.I wouldbe
disingenuousif I did not note that I have some ideas on how at least
one theory of their truthconditionsshould go, ideas that take seriously
the agency of the agent. These ideas are not all that rich, but they at
least suffice to suggest a possible explanationas to why there are just
four modes of action (and of course four correlativemodes of inaction,
by negation).
seeingto it that Q:[a stit:Q]
seeingto it that - Q: [a stit: - Q]
seeingto it that Q: [a stit: - [a stit:Q]],and
refrainingffrom
seeingto it that - Q:[a stit: - [a stit: - Q].
refrainingffrom
For example,the suggestedtruth-conditionalsemanticsexplainwhy, at
leastin the contextof certainrestrictions,
refrainingfrom refrainingfrom seeing to it that Q, that is,
[a stit: - [a stit: - [a stit:Qfl]
is not a new mode of action. And here is a furtherexercise:take the
standardreadingof the action of giving an order as seeing to it that
26
NUEL BELNAP
there is an obligationlaid on the addressedagent.Then the givingof an
order has the followingcontent, expressedsomewhatredundantly36so
as to be ableto call attentionto some possibleblanks:
[-I [a
stit:
2[a stit:
30bligated[F stit:-4 [P stit:-5Q11111
In each of the numbered blanks there is room for a negation. The
exercise is to see how instructiveit is to fill these blanks in various
ways, thus pulling together into a single schema some otherwise
confusing observations.For example, filling 3 and 5 is the positive
grantingof permission or authorization,while filling 2 and 5 is the
positive act of refraining from laying on an obligation, which is
sometimes thought of as a kind of permission,while filling 1 and 5
describesthe non-act of not laying on an obligation,which some also
think of as fallingwithinthe precinctsof permission.Sans scribed stitsentences, thyself the troubledtempted thinkertemerouslytumblesto
thrasonical tergiversantturpitude, treacherouslythroppling the true
track thirlingthroughthese turbid teeming twizzles;the tongue twists
too trickilytrippinglyto terminethese thorny termlesstwirlerstristily.
So scrupulouslyscribestits.
Stit-sentencesaside, hoever, here is a deceptively simple-sounding
question to which I, at least, have not heard a careful and helpful
answer:what is the meaningof a precative,that is, of a request?If you
prefer to answerin terms of performativesor in terms of illocutionary
force, fine: how is the world differentafter you have requestedyour
boss to give you a raise, or Jack to explain sincerityconditions?Even
though I do not have a theory about the matter, I will tell you
somethingI think might be true: requests are de re ties that bind us
person to person in the mode of caring, and are perhaps more
necessaryto undergirdmoralityand our life togetherthan even those
illocutionaryactsthatcreatecommitmentsor obligationsor rights.
3.3. Pragmaticsof Imperatives
Allow me to conveyjust one unifyingidea concerningthe pragmaticsof
sentences,whetherthey be imperatives,interrogatives,or even declaratives. Addressing the pragmaticsof interrogatives,I adumbratedthe
idea of picturingspeakersas carryinga slate listing their commitments
DECLARATIVES ARE NOT ENOUGH
27
(as membersof a linguisticcommunity,relativeto a specific situation),
and I pointed out that the slate must not be restrictedto lists whose
contentis declarative.Now I want to stress that the slate'scontentmust
not even be restrictedto declarativesor interrogatives.For example,
having mentionedcommitments,we need to insist that there ought to
be on that slate at least a list of commitmentsof a whollydifferentkind
from propositional commitments, namely, commitments to action,
which we can happilyrepresentwith stit-sentences[I stit:Q]. But what
we cannot do is representthese commitmentsto action by any conceivable propositionalcommitment;to suppose that we can is the last
dyinggaspof the DeclarativeFallacy.For example,
Jack'scommitmentto explainsincerityconditionsto Alfred
is to be representedon the action-commitmentportion of his slate by
the stit-sentence
[Istit:I explainsincerityconditionsto Alfred].
WhatI am sayingis that there is no propositionthat you can put on the
list of propositionsto which he is committedthat will do the representationalwork required.Furthermore,since every normativeand interpersonal structureinvolves the creation of commitmentsto action, no
contextualaccountof anythinghumanis possible that does not at least
secretly rely on stit-sentences.Let us togetherassume the philosophic
taskof makingthe secretpublic.
I hope to have convincedyou that these three lists differnot just in
what they representas lists, but more profoundly,that they differin the
form of theircontent,and that they are interdependent.Along the way I
hope to have emphasizedthat in developingour philosophicalunderstanding of topics touching on these contents we should include
theoriesof grammar,of compositionalsemantics,of speech acts, and of
largernormativeor interpersonalstructures.And converselyand above
all, I hope to have emphasized that every philosophical program
touchinglanguagecan profit from recognitionof the separaterequirements and importanceof declaratives,interrogatives,and imperatives.
Each of these three forms of content37must be conceived as being all
on the same slate, all interdependent;to do philosophicaljustice to any
willrequiredoingjusticeto all.
28
NUEL BELNAP
NOTES
1 Thanksare due to the Universityof PittsburghCenterfor the Philosophyof Science,
under the sponsorshipof which I read the antecedentof this paper in Februaryof
1988. I am profoundlyindebted to J. Seibt for knowledgeable,talented, and timeconsumingsubstantivecontributionsto this paper throughoutthe entire period of its
composition.
2 Some speech act theoristsuse 'declarative'
as part of their technicalterminologywith
a differentmeaning,to signify the sort of act in which we make somethingtrue by
declaringit so. Here, however, 'declarative'is used to help make the standardgrammaticalcontrastwithinterrogativesandimperatives.
I See G. Frege, Foundationsof Arithmetic,tr. by J. L. Austin, Oxford,Basil Blackwell,
1959, sect. 60, 62, pp. 71e, 73.
4 This is certainlythe way that contemporariesunderstandhim. See M. Dummett,
'What is a theory of meaning? (II)', in: G. Evans/J. McDowell (eds.), Truth and
Meaning,Oxford,ClarendonPress, 1981, ch. 19, pp. 428-472; or McDowell,'Truth
conditions, bivalence, and verificationism',G. Evans/J. McDowell, op. cit., p. 44.
Frege's position is that interrogativesand imperatives,unlike declaratives,lack a
reference.See, for instance,Frege, 'On Sense and Reference',in: P. Geach/M. Black
(eds.), TranslationsFrom the Philosophical Writingsof Gottlob Frege, Oxford, B.
Blackwell,1960, p. 68.
5 See J. Searle/D. Vanderveken,Foundationsof illocutionarylogic, Cambridge,CambridgeUniversityPress, 1985.
6 See M. Pendlebury,'Against the Power of Force: Reflections on the Meaning of
Mood',Mind,vol. xcv, 1986, pp. 361-372.
7 I also have to make good on an impliedpromise to distinguishand relate the three
differentcategoriesfor which I use the adjective'imperative':the grammaticalcategory
of imperativesentences(stand-aloneand constituent),the abstractcategoryof imperative contents, and the pragmaticcategory of imperativeacts. But this is beyond the
scope of this paper.
8 See D. Davidson,'TruthandMeaning',Synthese,vol. 17, 1967, pp. 304-323; reprinted
in:Inquiriesinto Truthand Interpretation,
Oxford,ClarendonPress, 1984, pp. 17-37.
1 See M. Bennett, Questionsin Montague-Grammar,
Bloomington,IN, IndianaLinguisticsClub,1979.
10 See e.g.Dummett,op. cit.
" See A. Ross, 'Imperativesand Logic', Theoria,vol. 17, 1941, pp. 53-71; also in:
Philosophyof Science,vol. 11, 1944, pp. 30-46.
12 See B. Chellas, The Logical Form of Imperatives,Stanford,CA, Perry Lane Press,
1969.
13 Of course you can give a reasonfor the true answerto a questionbeingwhichone it
is. Furthermore,reason why is a special case, obtainingbecause why-questionsthemselves ask for reasons.To give a reasonwhy truthconditionscause canceris just to give
an answer to the question of why truth conditions cause cancer; it is not to give a
reasonfor the questionitself of why truthconditionscause cancer- that appearsto be
nonsense.
14 See C. Hamblin,Imperatives,
New York,B. Blackwell,1987.
5 The best introductionto the theory of interrogativesis D. Harrah,"The Logic of
Questions,"Handbook of PhilosophicalLogic, vol. II: Extensionsof ClassicalLogic,
D. ReidelPub.Co., 1984, pp. 715-764.
Dordrecht-Boston-Lancaster,
16 Elisabet Engdahl calls them 'constituentquestions'in Engdahl, ConstituentQuestions,Dordrecht-Boston,D. ReidelPub.Co., 1986.
17 In fact, not only some who-clausesbut also some if-clauses cannot satisfyinglybe
treatedwithinthe grammarof declaratives.J. L. Austin in his celebratedpaper 'Ifs and
DECLARATIVES ARE NOT ENOUGH
29
Cans' (see Austin, 'Ifs and Cans', in: PhilosophicalPapers, Oxford, ClarendonPress,
1961, pp. 153-181) pointed out that in a sentence like I can if I choose, the if-clause
cannot be taken as expressinga conditional.But while this negative observationof
Austin's found ample response, nobody seems to have even reacted,neither affirmativelynor critically,to his positive thesis (see op. cit.,p. 160) that the clause'if I choose'
expresses a question!The fact that we 'wonder'or 'ask' if such and such is the case
supportsthe plausibilityof Austin'sconjectureand should make us wonderor ask why
philosophersstill indulge in declarativeslumber.For a discussionof embeddedinterrogativeswith if-clauses see Bolinger, 'Yes-No Questions Are Not AlternativeQuestions', in: H. Hiz (ed.), Questions,Dordrecht-Boston,D. Reidel Pub. Co., 1978, pp.
87-107.
18 See Hamblin,'Questions',TheAustralasian
journal of philosophy,vol. 36, 1958, pp.
159-168.
19 See B. Brandom,'Asserting',Nous, vol. 17, 1983, pp. 637-651.
20
See Hamblin,Imperatives,New York,B. Blackwell,1987, p. 229.
21 See R. G. Collingwood,An Essay in Metaphysics,Oxford, Clarendon,1962, p. 23,
and the detailedworkingout in R. Manor,'Pragmaticsand the Logic of Questionsand
Assertions',Philosophica,vol. 29, 1982, pp. 45-95.
22 See for instanceD. Harrah,Communication:
a logical model, CambridgeMA, MIT
Press, 1963.
23 The best introductionto the theory of imperativesis Hamblin'spreviously cited
book, Imperatives.For a penetratingdiscussionthat avoids the DeclarativeFallacyand
may very well be combinablewith some suggestionsmade below, see M. Huntley,'The
Semanticsof EnglishImperatives',Linguistics&Philosophy,1984, pp. 103-133.
24 See H. N. Castanieda,Thinkingand Doing, Dordrecht-Boston,D. Reidel Publ. Co.,
1975,p. 169.
25 Those engaged in the descriptive grammarof English have and are entitled to
differentviews on this matter.Perhapsthe work most pertinentto our concerns is W.
Badecker,Formal Grammarsand the Analysis of Infinitives,Bloomington,IN, Indiana
UniversityLinguisticsClub, 1987. Badeckersurveyssome Chomskytheories,whichby
deriving all infinitive constructionsby transformationof declarativesare deeply at
variancewith the spirit of the present paper, thoughmy aims are so differentfrom his
thatit is hardto call the variancea conflict.In healthycontrast,the lexicalisttheorythat
Badecker offers in his chapter 3 awards infinitive constructionsindependencefrom
declaratives,and therebymore nearlysharesour direction;however,there remainsthe
questionof whetherin agentiveinfinitiveconstructionssuch as Jack refusedto explain
sincerityconditionsto Alfred,we should or should not take it that there is a "trace"of
Jack headingthe infinitivephrase. We certainlyneed Jack to get the semanticsright,
but that far from settles the syntacticalquestion for English. In any event, Badecker
suppliesa trulyhelpfulframeworkfor addressingthis andrelatedquestions.
26 See A. R.
Anderson,'Logic,normsand roles',Ratio,vol. 4, 1962, pp. 36-49.
27 In particular,
it just doesn'tmatterwhetheror not Q is itselfagentivein a.
28 I amignoringtense as a temporaryanddangerousstrategy.
29 On my proposal,if Q is agentive,then
la stit: Ql is equivalentto Q. Obviouslythis is
not to be taken as an analysis of Q, especially not as an analysis of Q in terms of
propositionsthat are non-agentive.In this way the programis very differentfrom that
of say Chisholm. See R. Chisholm, 'Evidence as Justification',The Journal of Philosophy, vol. 58, 1961, pp. 739-748, or 'The Ethics of Requirement',American
Philosophical Quarterly,vol. 1, 1964, pp. 147-154; or A. Kenny, Action, Emotion
and Will,London,Routledge& KeaganPaul, 1963, ch. 8; both of these programsinsist
on replacingQ with a sentencedescribingsome terminalstate of affairs.In contrastwe
clarifybutdo not analyze.
310 Note: subjunctive,not indicative,as in Mary demandedthat Jack explain sincerity
conditions.
30
NUEL BELNAP
I'mnot sure this use of overlappinggrammaticalcategoriesis the best way to go. Let
me emphasizeagain that the aim of applyingstit-notationis to clarify,not to analyse
and, in particular,not to providea syntacticalcriterionfor when a certainsurfaceform
mustbe consideredan imperativeratherthana declarative.
32 This equivalenceis fine for agentive Q; thus, to be forbidden to see to it that you
explainsincerityconditionsto Alfred is indeed equivalentto being obligatedto see to it
that you do not explain sincerityconditions to Alfred, precisely because you explain
sincerityconditions to Alfred is agentive.This remarkis not ad hoc, but a straightforward,useful, and importantconsequenceof the analysis.Dialecticallywe use it to
explainhow easy it is to confuse these mattersand thereforehow easy it has been for
some (but far from all) investigators,trainedin the DeclarativeFallacy, to get things
wrong.
33 See Castanieda,
op. cit.
34 See for instanceG. H. v. Wright,'DeonticLogic',Mind,1951, pp. 1-15.
35 See Hamblin,op. cit.
36 [a stit: [a stit: Q]lcollapses into [a stit: QJ.
17 And as few othersas possible.
31
Departmentof Philosophy,
Facultyof Artsand Sciences,
Universityof Pittsburgh,
Pittsburgh,PA 15260,
U.S.A.