Você está na página 1de 6

Royal Institute of Philosophy

Hume's Law and Hare's rule


Author(s): Amartya K. Sen
Source: Philosophy, Vol. 41, No. 155 (Jan., 1966), pp. 75-79
Published by: Cambridge University Press on behalf of Royal Institute of Philosophy
Stable URL: http://www.jstor.org/stable/3749039 .
Accessed: 31/07/2014 05:28

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Cambridge University Press and Royal Institute of Philosophy are collaborating with JSTOR to digitize,
preserve and extend access to Philosophy.

http://www.jstor.org

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:28:00 AM


All use subject to JSTOR Terms and Conditions
DISCUSSIONS

no falsifiableclaims.We muston no accountoverlookthata theorymay


be in practiceunfalsifiablesimplybecause it happensin factto be true!
And that thisis so withthe centralcore of Darwinismis neithera fault
nor even a peculiarity.
5. One finalpoint.It would be valuable here,and oftenelsewheretoo,
whichMansernoticesbut failsto put to work.
to insistupon a distinction
This is that betweenthe questionof the evolutionof species and the
questionof the mechanismsresponsibleforsuch evolution:'forstrictlya
"theoryof evolution"should onlystatethatlivingcreatureshave come
intobeingas theresultofchangesin earlierforms... sucha theorycan be
completelyneutralas to how such changes take place. . . . Darwin's
importantcontribution was to suggesta mechanismof such changesin
we can saythat
thecase ofspecies'(pp. 19-20). In termsofthisdistinction
Special Creationwould be a rival to a theoryof evolutionin the strict
whichis scientific-
sense;and to sucha theorytheresurelyis no alternative
ally tolerable.Lamarckianismis not a rival or an alternativeeitherto
evolutionin this sense or to Special Creation. It offersa suggestion,
apparentlydisproved,about anothertheoretically possiblemechanismof
evolution.What Darwin did was to show that one immenselypowerful
powerfulmechanism,thatofnaturalselection,not merelyis but mustbe
operating;and therebyto providean enormously strongwarrantforthe
convictionthatevolutionis the truth,and thata scienceof evolutionary
biologyis possible.'
ofKeele.
University

3I wishto acknowledgeherethehelp givenme bymycolleagueDr K. Goodway


of the Departmentof Biologyin the Universityof Keele. He is, however,not
responsibleformymistakes.

HUME'S LAW AND HARE'S RULE


AMARTYA K. SEN

THIS note claims that contraryto his denial,' Mr Hare's 'adherence'to


'Hume's Law' conflictswithhis adherenceto 'universalprescriptivism'.
Or, to putit differently then
ifvaluejudgmentsare indeeduniversalisable,
thereis at least one value judgmentthatfollowsfromexclusivelyfactual
premises.
To keep theargumentsimple,we shall takea case ofa nonmoralvalue
judgmentdiscussedby Hare, where the patternof universalisability is
obvious.Considerthefollowing twostatements. one,
A purelydescriptive
to be called C: 'This motor-caris exactlylike the nextmotor-car'.The
second,to be called V, is a value judgment:'The nextcar is as good a
motor-caras thisone'. Examinenow the followingstatementofHare:
1R. M. Hare, Freedom and Reason(hereafterF.R.) (Oxford, 1963), pp. 108,
116, 186. Also his The Languageof Morals (hereafterL.M.) (Oxford, 1961),
pp. 30-31.

75

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:28:00 AM


All use subject to JSTOR Terms and Conditions
PHILOSOPHY

'I cannot say "This is a good motor-car,but the one next to it, though
exactlylikeit in all otherrespectsis notgood".' (L.M., p. 135.)
Does this violate Hume's Law? Not directly,forwhat is being claimed
here is that if I do make a value judgment, which we may call V* (that
this is a good motor-car), and accept a factual statement,which we may
call C (that the next one is exactly like it in all descriptivefeatures), then
I cannot make another value judgment, V** (that the next motor-car is
not good).
We can look at the argument slightlydifferently.If the firstcar is good
and the next car cannot be called 'not good', what is being shown here is
that I cannot deny that the next car is as good a motor-caras this one, i.e.
I cannot deny V, as defined above. So it is being claimed that I cannot
say, 'V*, C, and notV'. Would Hare permit us to drop out V* fromthis,
and say 'C and notV'? Can we say, forexample: 'This is a medium quality
car. The one next to it, though exactly like this one in all other respects
is not as good as this one'? Clearly, according to Hare we cannot say this
either, because good being a 'supervenient epithet', two objects cannot
differonly in 'goodness' being exactly alike (L.M., pp. 80-81, 130-1). So
V* is really inconsequential in this context. What, according to Hare,
we cannot say is simply: 'C and notV'.
In truth-functionallogic, to deny 'C and notV' is to assert 'C implies
V'. But this inferencecannot be immediately applied to the logic of value
statements,as it might be Hare's intention to claim that we cannot say
'C and notV' without committingus to saying 'if C then V'. He might
mean to suggest that if we say C, we cannot say 'notV', but we need not
affirmV either. However, it is easily confirmedthat this is not what Hare
is suggesting; he is saying that we are committed to affirming V in such a
situation. 'And if I call a thing a good X, I am committed to calling any
X like it good.' (F.R., p. 15.) This settlesone problem but once again this
looks like an argument with av alue premise (this is a good X) and a
factual premise (another X is exactly like it) with a value conclusion (the
other X is also a good X). But it is easy to show (just as in the last para-
graph) that what is being asserted is 'if C then V' (substitutingX for
motor-cars),i.e. if I agree that A is like B as an X, then I am committed
to saying that A is as good an X as B.
We arrive, therefore,at the position that according to Hare ifwe do say
C, not only can we not say 'not V', but we are committed to saying V.
But this means that there is a value judgment that can be derived froman
exclusively factual premise. This is not transparent because superficially
the argument has the form: 'if V * and C, then V * *', when V * stands for
the value judgment 'A is a good X', C forthe factthat A and B are exactly
alike, and V** another value judgment 'B is a good X'. To see how this
violates Hume's Law, we can go in two steps. First we convert the argu-
ment into affirming:'if C, then V* implies V**', i.e. 'if A and B are
descriptivelyalike, then A being good implies B being good'. This itselfis
a violation of Hume's Law, but since the conclusion is in the hypothetical
form,this might look like a trivial violation of it. (In fact,it is nota trivial
76

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:28:00 AM


All use subject to JSTOR Terms and Conditions
DISCUSSIONS

violation;contrastthe much-discussed trivialcase: 'if x theny' implies


'ify oughtto be avoided thanx oughtto be avoided'.) We can, however,
replace'good' by 'mediumquality','poor', 'highclass',or anyothervalue
expression;and consistentwith 'universalisability' of all these value
judgments,Hare will confirmthe statementin each case. So that the
conclusionis reallymore generalthan 'A being good impliesB being
good'. It is oftheform'A is as good as B', a relativeevaluationofthetwo,
irrespective of the absolutevalue of A's goodness.Thus the argumentis
indeedone, as claimedearlier,represented by 'if C thenV'.
The sourceof thisproblemis this.Hare defines'good' and 'better'in
termsof 'ought' (L.M., section12.3), and 'ought'in termsof an implied
imperative(L.M., pp. 168-9); thisfitsin with'prescriptivism'. Butbecause
of 'universalism',the notionof 'goodness'as studiedby Hare cannotbe
independentofdescriptive features.Hare statesthisquite clearlyhimself,
but pointsout thatthereis no uniquerelationbetweendescriptive features
and their'goodness',and criticisesthe naturalistsfortyingvalue judg-
ments 'analyticallyto a certaincontent' (F.R., p. 195). But thanksto
universalism he has to acceptthatiftwoobjectshave thesame descriptive
features,they cannot differin goodness; and even this modest claim
violatesHume's Law. Hare seemsto overlookthisbecause he poses the
problemin themodelof'ifV* and C, thenV* *', as discussedearlier.But
thanksto universalisability of all value judgments,as shownabove, this
amountsto claiming'ifC, thenV', whenC is a factualstatement and V a
nontrivialvalue judgment.
Hare devotesa lot oftimeto distinguishing hispositionfromthatofthe
naturalists. In thiscontext,he pointsout that'fora naturalisttheinference
froma non-moraldescriptionof somethingto a moral conclusionabout
it is an inference whosevalidityis due solelyto themeaningof thewords
in it'. (F.R., p. 21.) Acceptingthisdescription ofa naturalist(and bearing
in mindthatthe pointat issuehereis not the distinction betweenmoral
and otherkindsofjudgments), we can definea naturalist as one whoclaims
that the inferencefroma factualstatementto a value judgmentis 'due
solelyto themeaningofthewordsin it'. We can distinguish betweentwo
kindsofnaturalists:'existentialnaturalists', thosewho claim thatat least
one value inferencecan be made like this,and 'universalnaturalists',
viz. thosewho claim thatall value inferences have thisproperty.What is
beingclaimedhereis thatHare's positionis an existentialnaturalistone.
The violationof Hume's Law followsfrom'the way in whichthe word
"good" functions'(L.M., p. 130).
The pointcan be put a littlemorepreciselyusingelementary concepts
ofset theory.Considerthethreefollowing sets,S. T, and U. Each object
to be comparedis givena different numberto identify, and thesetofthese
numbers,each standingfor one object, is S. Each object has certain
descriptivefeatures.The set of all possiblecombinationsof descriptive
features, we call T. Finally,'goodness'is represented by a setof numbers
U, such thata highernumberrepresents 'better'.(These numbersneed
not have cardinal properties.)In termsof 'ought' statements, thiscor-
77

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:28:00 AM


All use subject to JSTOR Terms and Conditions
PHILOSOPHY

respondsto sayingthatifone has to choosebetweentwoobjectswithtwo


different 'goodnessnumbers',and one cannothave both,thenone ought
to choosethe objectwiththe highernumber.
The universalnaturalistwillclaimthatby virtueofthemeaningofthe
words,thereis a unique'transformation' ofT intoU. If someonedoes not
agree thatone collectiona ofdescriptive features(i.e. one elementin T)
mustbe relatedto one particularvalue P ofgoodness(i.e. one elementin
U) givenby thisunique transformation, thenhe is revealinga cognitive
defect.Hare doesnotclaimthisand doesnot,therefore, advocateuniversal
naturalism.But his 'universalprescriptivism' does assertthatwhile each
of us can entertaina different transformation fromT to U, the relation
betweenthesetT and thesetU mustbe, foreach ofus, one oftransforma-
tion(in the strictsense)fromT to U. That is foreach elementin T there
is (foreach of us) oneandonlyoneelementin U (thoughforeach element
in U theremightbe more than one elementin T). This means that
ifwe take two objects,i.e. two numbersfromthe set S, but iftheycor-
respondto thesame elementin T, i.e. have thesame descriptive features,
then theymustcorrespondto the same elementin U, i.e. be 'equally
good', or one 'as good as' the other.
To illustrate,in the old example,the firstcar and the one next to it
representedtwo different elementsin the set S, but having the same
features,correspondedto the same elementin T, and thus, thanksto
universalism,had the same goodnessnumber,or were equally good.
Thus, while universalnaturalismrequiresthat therebe a unique trans-
formationof T into U, Hare's positionrequiresonly that each relation
betweenT and U mustbe one ofa transformation ofT intoU, thoughnot
necessarilya unique one sharedby everybody.But thisviolatesHume's
Law, becausetwoelementsin S (twomotor-cars), becausetheycorrespond
to the same elementin T (have the same descriptivefeatures),mustbe
judged to correspondto the same elementin U (be regardedas good as
each other).And it is in thissensethatHare's positionis an existential
naturalistone.
We shouldnow brieflyreferto the backgroundquestionwhethertwo
separate objects (or situations)can be exactly alike. For a varietyof
reasonsthisproblemis not a fundamentalone forour purpose.Firstly,
forHare's universalisability, theyneed not actually be exactly alike; if
theyare thoughtto be so, thatis sufficient. Secondly,Hare also considers
cases wheretwo objectsdifferin some respectsthat are irrelevantto the
choice (F.R., pp. 140-41). Provided they are exactly alike in other
respects,universalisability is applied. The notionofrelevancebringsin a
possibleambiguity,in case thereis disagreement about what is relevant.
But thisdoes not make Hume's Law any morevalid. Note the following
argumentto which Hare is committed(L.M., pp. 129-30; F.R., pp.
140-41): 'A and B have thesame descriptive featuresexceptin respectR'
implies 'If respectR is irrelevantto the choice,thenA and B are equally
good'. The conclusionis a conditionalvalue judgmentdependingon
anothervalue judgmentimpliedin the notionofrelevance,but the con-
78

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:28:00 AM


All use subject to JSTOR Terms and Conditions
DISCUSSIONS

ditionis such a weak one thatit does not make the conclusiontrivialin
any sense.(It is likeconcludingin themotor-carargument:iftherelative
situationofthiscar and thenextone on theshopflooris irrelevant to their
relativegoodness,and theyare exactlyalike otherwise,then they are
equally good.) The forceofthisconditionaljudgmentis hardlyless than
that of V in our exact-likenesscase. Lastly,we need not reallyspend
sleeplessnightson whetherA and B can be relevantlyalike. If they
cannot,then Hare's principleof universalisability is emptyof content.
If theycan, thenmorethanone elementin S can correspondto thesame
elementin T, and theproblemdiscussedherearises.A moreformalstate-
mentof the propositionbeing provedhere is: EitherHare's principleof
universalisability is emptyof content,orit conflicts
withHume's Law.
I shouldemphasisethatI am notarguinghereagainstHare's 'universal
prescriptivism' whichis certainlyamongthe mostfruitful approachesto
ethicaldiscussions.I am onlyobjectingto combiningthisapproachwith
a claim ofstrictadherenceto Hume's celebrated'law'. It does notworry
me unduly to think that Hare's universalprescriptivism implies an
'existentialnaturalist'position,but it worriesme to thinkthat this
implicationis denied.

University at Berkeley.
ofCalifornia,

MR STRAWSON ON THE NOTION OF 'PREDICATE'

M. DURRANT

IN his discussionof 'Subject and Predicate' (Individuals-AnEssay in


Descriptive Metaphysics,p. 139 ff.),Mr P. F. Strawsonpresentsus with
listsof expressionsA1 B1 which he claims are listsof expressionswhich
philosophershave used to expressa fundamentalfunctionaldistinction
between'pickingout an individual'and 'sayingsomethingabout it' as
between'pickingout a subject'and 'sayingsomethingabout it'.
I shall concentrateon one elementin this distinction-viz.that of
'sayingsomethingabout a subject' and firstdiscusswhetherit is not a
misunderstanding of our language to thinkthatwhenevera sentenceof
grammaticalsubject-predicate formis used to make a statement, thereis,
as faras thepredicateis concerned,basicallyjust one thingbeingdone-
viz. thatofsayingsomething about a subject.I shallfurther be concerned
to pointout thatMr Strawson's'categorycriterion'forthe distinction of
subjectand predicateis inadequate, and finallyI shall argue that it is
misleadingto talkof'two thingsbeingdone in makinga statement'-viz.
pickingout a subjectand predicatingsomethingofit.
First,it would be incorrectto say thatthe notionthat all expressions
which can grammatically occur as predicatesdo one thingbasically,is
arrivedat merelyby givinga generalanswerto the generalquestion:
'What do all expressions whichcan occuras grammaticalpredicatesdo?'
That theyall 'say something about a subject'is notjust an answerto this
79

This content downloaded from 119.15.93.148 on Thu, 31 Jul 2014 05:28:00 AM


All use subject to JSTOR Terms and Conditions

Você também pode gostar