1.new Work For A Theory of Universals PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 36

PLEASE SCROLL DOWN FOR ARTICLE

This article was downloaded by: [Columbia University]


On: 29 September 2009
Access details: Access Details: [subscription number 908759819]
Publisher Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Australasian Journal of Philosophy
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713659165
New work for a theory of universals
David Lewis
a
a
Princeton University,
Online Publication Date: 01 December 1983
To cite this Article Lewis, David(1983)'New work for a theory of universals',Australasian Journal of Philosophy,61:4,343 377
To link to this Article: DOI: 10.1080/00048408312341131
URL: http://dx.doi.org/10.1080/00048408312341131
Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Australasian Journal o f Phi l osophy
Vol. 61, No. 4, December 1983
NE W WOR K F OR A T HE OR Y OF UNI VE R S AL S
Da v i d Le wi s
I nt roduct i on. D. M. Ar ms t r o n g of f e r s a t h e o r y o f uni ve r s a l s as t he o n l y
adequat e a n s we r t o a ' c o mp u l s o r y q u e s t i o n ' f or s y s t e ma t i c p h i l o s o p h y : t he
pr obl em o f On e ove r Ma ny. 1 I f i nd t hi s l i ne o f a r g u me n t u n p e r s u a s v i e . But
I t hi nk t he r e is mo r e t o be s a i d f or Ar ms t r o n g ' s t h e o r y t h a n he h i ms e l f ha s
said. Fo r as I be a r i t i n mi n d cons i der i ng va r i ous t opi c s i n p h i l o s o p h y , I not i ce
time a n d a g a i n t h a t i t o f f e r s s o l u t i o n s t o my p r o b l e ms . Wh a t e v e r we ma y
think o f t he p r o b l e m o f On e over Ma n y , uni ve r s a l s c a n e a r n t he i r l i vi ng d o i n g
other mu c h - n e e d e d wo r k .
I do n o t s a y t h a t t h e y a r e i n d i s p e n s a b l e . Th e s er vi ces t h e y r e n d e r c o u l d
be ma t c h e d us i ng r e s o u r c e s t h a t a r e No mi n a l i s t i c i n l e t t e r , i f p e r h a p s n o t
in spi r i t . 2 But n e i t h e r d o I h o l d a n y p r e s u mp t i o n a g a i n s t u n i v e r s a l s , t o t h e
effect t ha t t h e y a r e t o b e a c c e p t e d o n l y i f we h a v e n o a l t e r n a t i v e . I t h e r e f o r e
suspend j u d g e me n t a b o u t uni ve r s a l s t he ms e l ve s . I o n l y i ns i s t t h a t , one wa y
or a n o t h e r , t he i r wo r k mu s t be d o n e .
I shal l i nve s t i ga t e t he benef i t s o f a d d i n g u n i v e r s a l s t o my o wn us ua l
ont ol ogy. Th a t o n t o l o g y , t h o u g h No mi n a l i s t i c , is i n o t h e r r e s pe c t s ge ne r ous .
It consi st s o f pos s i bi l i a - p a r t i c u l a r , i n d i v i d u a l t hi ngs , s o me o f whi c h
compr i se o u r a c t u a l wo r l d a n d o t h e r s o f wh i c h a r e u n a c t u a l i s e d 3 - - t o g e t h e r
with t he i t e r a t i v e h i e r a r c h y o f cl as s es bui l t u p f r o m t h e m. Th u s I a l r e a d y
have at my d i s p o s a l a t h e o r y o f p r o p e r t i e s as cl asses o f p o s s i b i l i a . Pr o p e r t i e s ,
so u n d e r s t o o d , a r e n o t mu c h l i ke uni ve r s a l s . No r c a n t he y, u n a i d e d , t a k e
over t he wo r k o f uni ve r s a l s . Ne ve r t he l e s s t h e y wi l l f i gur e i mp o r t a n t l y i n wh a t
D. M. Armstrong, Universals and Scientific Realism (Cambridge University Press, 1978),
henceforth cited as "Universals" see also his "Against 'Ostrich' Nominalism: A Reply to Michael
Devitt", Pacific Philosophical Quarterly 61 (1980) pp. 440-449.
z In this paper, I follow Armstrong's traditional terminology: 'universals' are repeatable entities,
wholly present wherever a particular instantiates them; 'Nominalism' is the rejection of such
entities. In the conflicting modern terminology of Harvard, classes count as 'universals' and
'Nominalism' is predominantly the rejection of classes. Confusion of the terminologies can
result in grave misunderstanding; see W. V. Quine, ' Soft Impeachment Disowned', Pacific
Philosophical Quarterly 61 (1980) pp. 450-451.
A,laong 'things' I mean to include all the gerrymandered wholes and undemarcated parts
admitted by the most permissive sort of mereology. Further, I include such physical objects
as spatiotemporal regions and force fields, unless an eliminative reduction of them should
prove desirable. Further, I include such nonphysical objects as gods and spooks, though
not - I hope -- as parts of the same world as us. Worlds themselves need no special treatment.
They are things -- big ones, for the most part.
343
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
344 New Work f or a Theory of Universals
fol l ows, since f or me t hey ar e par t of t he envi r onment in whi ch universals
mi ght oper at e.
The fri end o f uni versal s ma y wonder whet her t hey woul d be better
empl oyed not as an addi t i on t o my ont ol ogy o f possi bi l i a and classes, but
r at her as a r epl acement f or par t s of it. A f ai r quest i on, and an ur gent one;
nevert hel ess, not a quest i on consi der ed in t hi s paper .
I n t he next sect i on, I shall sket ch Ar ms t r ong' s t heor y of universals,
cont rast i ng universals with propert i es under st ood as classes ofpossibilia. Then
I shall say why I a m unconvi nced by t he One over Ma n y ar gument . Then
I shall t ur n t o my pri nci pal t opi c: how universals coul d hel p me in connection
wi t h such t opi cs as dupl i cat i on, super veni ence, and di ver gent worl ds; a
mi ni mal f or m of mat eri al i sm; laws and causat i on; and the cont ent of language
and t hought . Pe r ha ps t he list coul d be ext ended.
Universals and Properties. Language of f er s us several mor e or less
i nt er changeabl e wor ds: ' uni ver sal ' ; ' pr oper t y' , ' qual i t y' , ' at t r i but e' , ' f eat ur e' ,
and ' char act er i st i c' ; ' t ype' , ' ki nd' , and ' sor t ' ; and per haps ot her s. And
phi l os ophy of f er s us several concept i ons of t he entities t hat such wor ds refer
t o. My pur pos e is not t o fix on one of t hese concept i ons; but r at her to
di st i ngui sh t wo (at oppos i t e ext r emes) and cont empl at e hel pi ng mys el f to
bot h. Ther ef or e s ome r egi ment at i on of l anguage is cal l ed f or ; I apologise
f or any i nconveni ence caused. Let me reserve t he wor d ' uni ver sal ' f or those
ent i t i es, i f such t here be, t hat mos t l y c onf or m t o Ar ms t r ong' s account . And
let me reserve t he wor d ' pr ope r t y' f or classes - any classes, but I have
f or e mos t in mi nd classes o f t hi ngs. To have a pr ope r t y is t o be a member
of t he class. 4
Why call t hem ' pr oper t i es ' as well as ' cl asses' ? - J us t t o under l i ne t he fact
t hat t hey need not be classes of actual t hi ngs. The pr oper t y of bei ng a donkey,
f or i nst ance, is t he class of all t he donkeys. Thi s pr ope r t y bel ongs t o - this
class cont ai ns - not onl y t he act ual donkeys of t hi s wor l d we live in, but
al so all t he unact ual i sed, ot her wor l dl y donkeys.
Li kewi se I reserve t he wor d ' r el at i on' f or ar bi t r ar y classes of or der ed pairs,
t ri pl es . . . . . Thus a r el at i on a mong t hi ngs is a pr ope r t y of ' t upl es of things.
Agai n, t here is no r est r i ct i on t o act ual t hi ngs. Cor r es pondi ng r oughl y to
t he di vi si on bet ween pr oper t i es and r el at i ons of t hi ngs, we have t he division
bet ween ' mona di c ' and ' pol yadi c' uni versal s.
Uni versal s and pr oper t i es di f f er in t wo pr i nci pal ways. The first difference
concer ns t hei r i nst ant i at i on. A uni ver sal is s uppos ed t o be whol l y present
wherever it is i nst ant i at ed. I t is a const i t uent par t (t hough not a spat i ot emporal
par t ) of each par t i cul ar t hat has it. A pr oper t y, by cont rast , is spr ead around.
The pr ope r t y of bei ng a donke y is par t l y pr esent wher ever t here is a donkey,
4 My conception of properties resembles the doctrine of Class Nominalism considered in
Universals, I, pp. 28-43. But, strictly speaking, a Class Nominalist would be someone who
claims to solve the One over Many problem simply by means of properties taken as classes,
and that is far from my intention.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 345
in this or a ny ot he r wor l d. Fa r f r o m t he p r o p e r t y bei ng pa r t o f t he d o n k e y ,
it is cl oser t o t he t r u t h t o s ay t ha t t he d o n k e y is p a r t o f t he pr ope r t y. But
the preci se t r ut h, r at her , is t ha t t he d o n k e y is a me mb e r o f t he pr ope r t y.
Thus uni ver sal s wo u l d u n i f y r eal i t y (Cf. Universals, I, p. 109) i n a wa y
that pr oper t i es do not . Thi ngs t ha t s har e a uni ver s al ha ve n o t j us t j oi ne d
a single class. Th e y l i t eral l y have s ome t hi ng i n c o mmo n . Th e y ar e n o t ent i r el y
distinct. Th e y over l ap.
By oc c ur r i ng r epeat edl y, uni ver sal s de f y i nt ui t i ve pr i nci pl es. But t ha t is
no da ma gi ng obj ect i on, since pl ai nl y t he i nt ui t i ons wer e ma d e f or par t i cul ar s.
For i nst ance, cal l t wo ent i t i es copresent i f b o t h ar e whol l y pr es ent at one
posi t i on i n s pace a nd t i me. We mi ght i nt ui t o f f h a n d t ha t c opr e s e nc e is
transitive. But it is n o t so, obvi ous l y, f o r uni ver sal s. Suppos e f o r t he sake
of a r gume nt t ha t t her e ar e uni ver sal s: r ound, silver, gol den. Silver and r o u n d
are copr es ent , f o r her e is a silver coi n; gol de n a n d r o u n d ar e copr es ent , f o r
there is a gol d coi n; b u t silver a n d gol de n ar e n o t c opr e s e nt . Li kewi se, i f
we add uni ver sal s t o an o n t o l o g y ofpossibilia, f or t he r el at i on o f bei ng pa r t
of t he s a me pos s i bl e wo r l d : I a n d s ome o t h e r wo r l d l y d r a g o n ar e not
wor l dmat es; b u t I a m a wo r l d ma t e o f t he uni ver s al gol de n, a nd so is t he
dragon. Pr e s u ma b l y I needed a mi xed case i nvol vi ng b o t h uni ver sal s a nd
part i cul ars. F o r wh y s houl d a n y t wo uni ver sal s ever fai l t o be wor l dma t e s ?
Lacki ng s uch f ai l ur es, t he wo r l d ma t e r el at i on a mo n g uni ver sal s al one is
ffivially t r ansi t i ve.
The s e c ond di f f er ence bet ween uni ver sal s a nd pr oper t i es c onc e r ns t hei r
abundance. Thi s is t he di f f er ence t ha t qual i fi es t h e m f or di f f e r e nt wo r k , a n d
t hereby gi ves rise t o my i nt er est i n ha vi ng uni ver sal s a nd pr oper t i es bot h.
A di st i nct i ve f eat ur e o f Ar ms t r o n g ' s t h e o r y is t ha t uni ver s al s ar e spar se.
There ar e t he uni ver sal s t ha t t her e mus t be t o g r o u n d t he obj ect i ve
resemblances and t he causal power s o f t hi ngs, a nd t her e is n o r eas on t o believe
in any mor e . Al l o f t he f ol l owi ng al l eged uni ver sal s wo u l d be r ej ect ed:
not gol den,
gol den or wo o d e n ,
met al l i c,
sel f-i dent i cal ,
owned b y Fr ed,
bel ongi ng t o cl ass C,
grue,
first e xa mi ne d b e f o r e 2000 A. D. ;
bei ng i dent i cal ,
bei ng al i ke i n s o me r espect ,
bei ng exact l y al i ke,
bei ng pa r t of ,
owni ng,
bei ng pa i r e d wi t h b y s ome pa r t i n R
If universals are to do the new work I have in store for them, they must be capable of repeated
occurrence not only within a world but also across worlds. They would then be an exception
to my usual principle - meant for particulars, of course - that nothing is wholly present
as part of two different worlds. But I see no harm in that. If two worlds are said to overlap
by having a coin in common, and if this coin is supposed to be wholly round in one world
and wholly octagonal in the other, I stubbornly ask what shape it is, and insist that shape
is not a relation to worlds. (See my 'Individuation by Acquaintance and by Stipulation',
PhilosophicalReview 92 (1983), pp. 3-32.) I do not see any parallel objection if worlds are
said to overlap by sharing a universal. What contingent, nonrelational property of the universal
could we put in place of shape of the coin in raising the problem? I cannot think of any.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
346
Ne w Work f o r a Theory o f Universals
(where C and R are ut t er l y mi scel l aneous classes). The gui di ng i dea, roughly,
is t hat t he wor l d' s uni ver sal s shoul d compr i s e a mi ni mal basi s for
char act er i si ng t he wor l d compl et el y. Uni ver sal s t hat do not cont r i but e at all
t o t hi s end ar e unwel come, and so ar e uni versal s t hat cont r i but e only
r edundant l y. A sat i sf act or y i nvent or y of uni versal s is a non-l i ngui st i c
count er par t o f a pr i mi t i ve voc a bul a r y f or a l anguage capabl e of describing
t he wor l d exhaust i vel y.
( That is r ough: Ar ms t r ong does not di smi ss r edundant uni versal s out of
hand, as t he spirit of his t heor y mi ght seem t o demand. Conj unct i ve universals
- as it mi ght be, gol den- and- r ound - ar e accept ed, t hough r edundant ; so
ar e anal ysabl e st r uct ur al uni versal s. The r eas on is t hat i f t he wor l d were
infinitely compl ex, t her e mi ght be no way t o cut down t o a mi ni mal basis.
The onl y al t er nat i ve t o r edundancy mi ght be i nadequacy, and i f so we had
bet t er t ol er at e r edundancy. But t he r edundancy is mi t i gat ed by t he f act t hat
compl ex uni versal s consist of t hei r si mpl er - - i f per haps not absol ut el y simple
- const i t uent s. They are not di st i nct ent i t i es. See Uni versal s, I I , pp. 30-42
and 67-71.)
I t is qui t e ot her wi se wi t h pr oper t i es. Any class o f t hi ngs, be it ever so
ger r ymander ed and mi scel l aneous and i ndescri babl e in t hought and language,
and be it ever so super f l uous in char act er i si ng t he wor l d, is nevert hel ess a
pr oper t y. So t her e are pr oper t i es in i mmens e abundance. ( I f t he numbe r of
t hi ngs, act ual and ot her wi se, is bet h- 2, an est i mat e I r egar d as mor e likely
l ow t han hi gh, t hen t he numbe r of pr oper t i es of t hi ngs is bet h-3. And t hat
is a bi g i nfi ni t y i ndeed, except t o st udent s of t he out er reaches o f set t heory. )
Ther e are so ma ny propert i es t hat t hose specifiable in English, or in t he brain' s
l anguage o f synapt i c i nt er connect i ons and neur al spi kes, coul d be onl y an
i nfi ni t esi mal mi nor i t y.
Because pr oper t i es are so a bunda nt , t hey are undi scr i mi nat i ng. Any two
t hi ngs shar e infinitely ma n y pr oper t i es, and fai l t o shar e infinitely many
ot her s. Tha t is so whet her t he t wo t hi ngs ar e per f ect dupl i cat es or utterly
di ssi mi l ar. Thus pr oper t i es do not hi ng t o capt ur e fact s of r esembl ance. That
is wor k mor e sui t ed t o t he sparse uni versal s. Li kewi se, pr oper t i es do nothing
t o capt ur e t he causal power s o f t hi ngs. Al mos t all pr oper t i es are causally
i r r el evant , and t her e is not hi ng t o ma ke t he r el evant ones st and out from
t he cr owd. Pr oper t i es car ve real i t y at t he j oi nt s - - and ever ywher e else as
well. I f i t ' s di st i nct i ons we want , t oo muc h st r uct ur e is no bet t er t han none.
I t woul d be ot her wi se i f we had not onl y t he count l ess t hr ong of all
pr oper t i es, but al so an 61ite mi nor i t y o f speci al pr oper t i es. Cal l t hese the
nat ur al pr oper t i es. 6 I f we had pr oper t i es and uni versal s bot h, t he universals
See Universals, I, pp. 38-41; Anthony Quinton, 'Properties and Classes', Proceedings of
the Aristotelian Society 48 (1957) pp. 33-58; and W. V. Quine, 'Natural Kinds', in his
Ontological Relativity (Columbia University Press, 1969). See also George Bealer, Quality
and Concept (Oxford University Press, 1982), especially pp. 9-10 and 177-187. Like me,
Bealer favours an inegalitarian twofold conception of properties: there are abundant 'concepts'
and sparse 'qualities', and the latter are the ones that 'determine t.he logical, causal, and
phenomenal order of reality'. (p. 10) Despite this point of agreement, however, Bealer's views
and mine differ in many ways.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Le wi s 347
could serve to pick out the natural properties. Afterwards the universals could
retire if they liked, and leave their jobs to the natural properties. Natural
properties would be the ones whose sharing makes for resemblance, and the
ones relevant to causal powers. Most simply, we could call a p r o p e r t y p e r f e c t l y
natural if its members are all and only those things that share some one
universal. But also we would have other less-than-perfectly natural properties,
made so by families of suitable related universals. 7 Thus we might have an
imperfectly natural propert y of being metallic, even if we had no such single
universal as metallic, in virtue of a close-knit family of genuine universals
one or another of which is instantiated by any metallic thing. These
imperfectly natural properties would be natural to varying degrees.
Let us say that an a d e q u a t e theory of properties is one that recognises an
objective difference between natural and unnatural properties; preferably,
a difference that admits of degree. A combined theory of properties and
universals is one sort of adequate theory of properties.
But not the only sort. A Nominalistic theory of properties could achieve
adequacy by other means. Instead of employing universals it could draw
primitive distinctions among particulars. Most simply, a Nominalist could
take it as a primitive fact that some classes of things are perfectly natural
properties; others are less-than-perfectly natural to various degrees; and most
are not at all natural. Such a Nominalist takes ' natural' as a primitive
predicate, and offers no analysis of what he means in predicating it of classes.
His intention is to select the very same classes as natural properties that the
user of universals would select. But he regards the universals as idle
machinery, fictitiously superimposed on the primative objective difference
between the natural properties and the others. 8
Alternatively, a Nominalist in pursuit of adequacy might prefer to rest with
primitive objective resemblance among things. (He might not think that
'natural' was a very natural primitive, perhaps because it is to be predicated
of classes.) Then he could undertake to define natural properties in terms
of the mutual resemblance of their members and the failure of resemblance
between their members and their non-members. Unfortunately, the project
meets with well-known technical difficulties. These can be solved, but at a
daunting price in complexity and artificiality of our primitive. We cannot
get by with the familiar dyadic ' resembles' . Instead we need a predicate of
resemblance that is bot h contrastive and variably polyadic. Something like
xl, x2 . . . . resemble one another and do not likewise resemble any of yl,
Y2, . . .
7 Here I assume that some solution to the problem of resemblance of universals is possible,
perhaps along the lines suggested by Armstrong in Universals, II, pp. 48-52 and 101-131;
and that such a solution could be carried over into a theory of resemblance of perfectly
natural properties, even if we take naturalness of properties as primitive.
8 This is the Moderate Class Nominalism considered in Universals, I, pp. 38-41. It is akin
to the view of Quinton, op. cit.; but plus the unactualised members of the natural classes,
and minus any hint that 'natural' could receive a psychologistic analysis.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
348 Ne w Wor k f or a The or y of Uni ver s al s
(where t he strings of vari abl es ma y be infinite, even uncount abl e) must be
t aken as under s t ood wi t hout f ur t her anaiysis. 9 I f adequat e Nomi nal i sm
requi res us t o choose bet ween this and a pri mi t i ve pr edi cat e of classes, we
mi ght well wonder whet her t he game is wor t h t he candl e. I onl y say we might
wonder ; I know of no consi der at i on t hat seems t o me decisive.
At t hi s poi nt , you may see ver y well why it coul d be a good i dea t o believe
in universals as well as pr oper t i es; but you ma y see no poi nt in having
pr oper t i es as well as uni versal s. But pr oper t i es have wor k of t hei r own, and
uni versal s are ill-suited t o do t he wor k of pr oper t i es.
It is pr oper t i es t hat we need, somet i mes nat ur al and somet i mes not , to
pr ovi de an adequat e suppl y of semant i c val ues f or linguistic expressions.
Consi der such sentences as these:
(1) Red resembl es or ange mor e t han it resembl es blue.
(2) Red is a col our .
(3) Humi l i t y is a vi rt ue.
(4) Redness is a sign o f ripeness.
Pr i ma f aci e, these sentences cont ai n names t hat cannot be t aken t o denote
par t i cul ar , i ndi vi dual t hi ngs. Wha t is t he semant i c rol e of t hese words? If
we are t o do composi t i onal semant i cs in t he way t hat is best devel oped, we
need entities t o assign as semant i c values t o t hese wor ds, entities t hat will
encode t hei r semant i c roles. Per haps somet i mes we mi ght find paraphrases
t hat will absol ve us f r om t he need t o subj ect t he original sent ence t o semantic
Such a t heor y is a f or m o f Resembl ance Nomi nal i s m, i n Ar ms t r ong' s cl assi fi cat i on, but it
is unl i ke t he f or m t hat he pr i nci pal l y consi ders. See Universals, I, pp. 44-63. For discussions
o f t he pr obl e m o f defi ni ng nat ur al cl asses in t er ms o f r esembl ance, and o f t he t r i cker y that
pr oves usef ul i n sol vi ng t hi s pr obl e m, see Nel son Goodma n, The St ruct ure of Appear ance
( Har var d Uni ver si t y Pr ess, 1951), Chapt er s IV-VI; W. V. Qui ne, ' Nat ur al Ki nds' ; and Adam
Mo r t o n , ' Compl ex I ndi vi dual s and Mul t i gr ade Rel at i ons' , NoDs 9 (1975) pp. 309-318.
To get f r om primitive resembl ance t o perfectly nat ural properties, I have in mi nd a definition
as f ol l ows. We begi n wi t h R as our cont r ast i ve and var i abl y pol yadi c pr i mi t i ve. We want
it t o t ur n out t hat x~,x2, . Ryl,Y2 . . . . i f f s ome per f ect l y nat ur al pr oper t y is shared by
all o f x~, x2, . but by none o f yl , Y2,. We want t o defi ne iV, anot her var i abl y polyadic
pr edi cat e, so t hat it will t ur n out t hat Nx bx 2 . . . . i f f x~, x2 . . . . are all and onl y t he members
o f s ome per f ect l y nat ur al pr oper t y. Agai n we mus t al l ow f or , and expect , t he case where
t her e ar e i nfi ni t el y many x' s. We defi ne Nxl , x2, . . . as:
3yl , y2 . . . . Yz (z, xl , x2, . . Rye, y2 . . . . ~ z =x ~ v z =x 2 v . . . ).
Then we finish t he j o b by defi ni ng a per f ect l y nat ur al pr oper t y as a class such t hat , i f xl,
x2 . . . . ar e all and onl y its member s , t hen Nx l , x2 . . . . .
We mi ght have t aken N as pr i mi t i ve i ns t ead o f R. But woul d t hat have been significantly
di f f er ent , gi ven t he i nt er def i nabi l i t y o f t he t wo? On t he ot her hand, t aki ng N as primitive
al so seems not si gni fi cant l y di f f er ent f r om t aki ng per f ect nat ur al ness o f classes as primitive.
It is onl y a di fference bet ween speaki ng i n t he plural of individuals and speaki ng in t he singular
o f t hei r cl asses, and t hat seems no real di f f er ence. Is pl ur al t al k a di sgui sed f or m of class
t al k? Or vi ce versa? (See t he di scussi on in Uni versal s, I, pp. 32-34; al so Max Black, ' The
El usi veness of Set s' , Re v i e w of Met aphys i cs 24 (1971) pp. 614-636; Er i c St eni us, 'Sets',
Synt hes e 27 (1974), pp. 161-188; and Kur t G6del , ' Russel l ' s Mat hemat i cal Logi c' , in P. A.
Schi l pp, ed. , The Phi l os ophy of Ber t r and Russel l ( Cambr i dge Uni ver si t y Pr ess, 1944).) At
any r at e, it is n o t at all cl ear t o me t hat Mode r a t e Cl ass Nomi nal i s m a nd Resemblance
Nomi nal i s m in its pr es ent f or m ar e t wo di f f er ent t heor i es, as oppos e d t o a single theory
pr es ent ed i n di f f er ent styles.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 349
analysis. That is t he case with (1), f or instance. I But even i f such paraphrases
sometimes exist -- even i f t hey always exist, whi ch seems unl i kel y - t hey
work pi ecemeal and f r ust r at e any syst emat i c appr oach t o semant i cs.
Ar mst r ong t akes it t hat such sentences pr ovi de a subsi di ary ar gument f or
universals, i ndependent of his mai n ar gument f r om t he One over Many
problem. (Universals, I, pp. 58-63; also "Agai nst ' Ost ri ch' Nomi nal i sm". n)
I quite agree t hat we have her e an ar gument f or somet hi ng. But not f or
universals as opposed t o pr oper t i es. Pr oper t i es can serve as t he requi si t e
semantic val ues. I ndeed, pr oper t i es are much bet t er suited t o t he j ob t han
universals are. That is plain even f r om t he examples consi dered. It is unl i kel y
that t her e are any such genui ne universals as t he col ours (especially
determinable col our s, like red, r at her t han det er mi nat e shades), or ripeness,
or humility. Ar mst r ong agrees (Universals, I, P. 61) t hat he cannot t ake (1)-(4)
as st r ai ght f or war dl y maki ng r ef er ence t o universals. He must first subj ect
them t o par aphr ase. Even i f t her e always is a par aphr ase t hat does r ef er t o,
or quant i f y over, genui ne uni versal s, still t he need f or par aphr ase is a t hr eat
to systematic sematics. The pr obl em arises exact l y because uni versal s are
sparse. Ther e is no cor r espondi ng obj ect i on i f we t ake t he requi si t e semant i c
values as propert i es.
Ot her sent ences make my poi nt mor e dramat i cal l y.
(5) Grueness does not make f or resembl ance among all its instances.
(6) What is c ommon t o all who suf f er pai n is bei ng in some or anot her
st at e t hat occupi es t he pai n rol e, pr esumabl y not t he same st at e in
all cases.
The poi nt is not t hat t hese sentences are t r ue -- t hough t hey are -- but t hat
they requi re semant i c analysis. (It is i rrel evant t hat t hey are not or di nar y
language.) A uni versal of grueness woul d be anat hema; as woul d a uni versal
such t hat , necessarily, one has it i f he is in some st at e or ot her t hat occupi es
10 In vi rt ue o f t he cl ose r es embl ance of r ed a nd or ange, it is possi bl e f or a r ed t hi ng t o r esembl e
an or ange one ver y cl osel y; it is n o t pos s i bl e f or a r ed t hi ng t o r es embl e a bl ue one qui t e
so closely. Gi ven our ont ol ogy ofpossibilia, all possibilities are realised. So we coul d par aphas e
(1) by
(1 ' ) Some r ed t hi ng r es embl es s o me or a nge t hi ng mo r e t h a n a n y r ed t hi ng r es embl es
a ny bl ue t hi ng.
so l ong as it is u n d e r s t o o d t ha t t he t hi ngs i n que s t i on ne e dn' t be par t o f our wor l d, or o f
any one wor l d. Or i f we di d not wi sh t o s peak o f una c t ua l i s e d t hi ngs , but we were wi l l i ng
to t ake or di na r y- l a ngua ge mo d a l i di oms as pr i mi t i ve, we coul d i ns t ead gi ve t he pa r a phr a s e :
( 1 " ) A red t hi ng c a n r esembl e a n or a nge t hi ng mo r e cl osel y t h a n a r ed t hi ng can r es embl e
a bl ue t hi ng.
It is necessar y t o us e t he or di nar y- l anguage i di oms , or s ome adequat e f or mal i s at i on o f t he m,
rather t h a n s t a nda r d mo d a l logic. You c a nnot expr ess ( 1' ' ) in moda l l ogi c ( excl udi ng a n
enriched moda l logic t hat woul d def eat t he poi nt o f t he par aphr as e by quant i f yi ng over degrees
of resembl ance or what not ) because you cannot express cross-worl d rel at i ons, a nd i n par t i cul ar
cannot expr ess t he ne e de d cr os s - wor l d c o mp a r i s o n o f si mi l ar i t y.
n He deri ves t h e a r g u me n t , a n d a s e c ond s e ma nt i c a r g u me n t t o be cons i der ed s hor t l y, f r o m
Ar t hur Pa p , ' No mi n a l i s m, Empi r i c i s m, a nd Uni ver s al s : I ' , Philosophical Quarterly 9 (1959)
pp. 330-340, a n d F. C. J a c ks on, ' St at ement s a b o u t Uni ver s al s ' , Mi nd 86 (1977) pp. 427-429.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
350 New Work for a Theory of Universals
t he pai n rol e in his case.lZ But t he cor r es pondi ng pr oper t i es ar e no pr obl em.
I ndeed, we have a compr ehens i on schema appl yi ng t o any pr edi cat e phrase
what ever , however compl i cat ed. (Let it even be infinitely l ong; let it even
i ncl ude i magi nar y names f or entities we haven' t real l y named. ) Let x range
over t hi ngs, P over pr oper t i es (classes) o f t hi ngs. Then:
31PV7x (x has P =-- l~x).
We coul d appr opr i at el y call t hi s ' t he pr ope r t y of It-ing' in t hose cases where
t he pr edi cat e phr ase is shor t enough t o f or m a ger und, and t ake t hi s pr oper t y
t o be t he semant i c val ue of t he ger und. Cont r as t t hi s wi t h t he ver y di fferent
r el at i onshi p of uni versal s and pr edi cat es set f or t h in Universals, I I , pp. 7-59.
Consi der al so t hose sent ences whi ch prima facie i nvol ve second- or der
quant i fi cat i on. Fr om Universals, I, p. 62, and "Agai nst ' Ost ri ch' Nomi nal i sm"
we have t hese.
(7) He has t he s ame vi r t ues as his f at her .
(8) The dresses were o f t he s ame col our .
(9) Ther e ar e undi scover ed f unda me nt a l physi cal pr oper t i es.
(10) Acqui r ed char act er i st i cs ar e never i nheri t ed.
(11) Some zool ogi cal species ar e cross-fert i l e.
Prima facie, we ar e quant i f yi ng ei t her over pr oper t i es or over universals.
Agai n, par aphr as es mi ght def eat t hat pr es umpt i on, but in a pi ecemeal way
t hat t hr eat ens syst emat i c semant i cs. I n each case, pr oper t i es coul d serve as
t he val ues of t he var i abl es of quant i fi cat i on. Onl y in case (9) coul d universals
serve equal l y well. To t r eat t he ot her cases, not t o ment i on
(12) Some charact eri st i cs, such as t he col our s, are mor e di sj unct i ve t han
t hey seem.
as quant i f i cat i ons over uni versal s, we woul d agai n have t o r esor t t o some
pr el i mi nar y par aphr as e. ( Ar ms t r ong agai n agrees: Universals, I, p. 63.) This
second semant i c ar gument , l i ke t he first, adduces wor k f or whi ch propert i es
ar e bet t er qual i fi ed t han uni versal s.
Whi ch is not t o deny t hat a par t ner s hi p mi ght do bet t er still. Let it be
gr ant ed t hat we ar e deal i ng wi t h quant i f i cat i ons over pr oper t i es. Still, these
quant i f i cat i ons - l i ke mos t o f our quant i f i cat i ons - ma y be t aci t l y or
expl i ci t l y rest ri ct ed. I n par t i cul ar , t hey usual l y a r e rest ri ct ed t o nat ural
pr oper t i es. Not t o per f ect l y nat ur al pr oper t i es t hat cor r es pond t o single
uni ver sal s, except in special cases like (9), but t o pr oper t i es t hat ar e at least
s omewhat mor e nat ur al t han t he gr eat maj or i t y of t he ut t er l y miscellaneous.
Tha t is so f or all our exampl es, even (12). Then even t hough we quant i fy
over pr oper t i es, we still need ei t her universals or t he resources of an adequat e
Nomi nal i s m in or der t o say whi ch o f t he pr oper t i es we mos t l y quant i f y over.
12 Or better, in the case of creatures of his kind. See my 'Mad Pain and Martian Pain', in Ned
Block, ed., Readings in Philosophy of Psychology, I (Harvard University Press, 1980).
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 351
I al so t hi nk t hat it is pr oper t i es t hat we need in char act er i si ng t he cont ent
of our i nt ent i onal at t i t udes. I bel i eve, or I desire, t hat I live in one o f t he
worlds in a cer t ai n class, r at her t han any wor l d out si de t hat class. Thi s class
of wor l ds is a pr ope r t y had by worl ds. I bel i eve, or I desi re, t hat my wor l d
has t hat pr oper t y. ( The class o f wor l ds al so ma y be cal l ed a proposi t i on,
in one of t he l egi t i mat e senses of t hat wor d, and my ' pr opos i t i onal at t i t ude'
of bel i ef or desi re has t hi s pr opos i t i on as its ' obj ect ' . ) Mor e general l y,
subsumi ng t he pr evi ous case, I bel i eve or I desi re t hat I mys el f bel ong t o
a cert ai n class of possi bi l i a. I ascri be a cert ai n pr ope r t y t o mysel f , or I want
to have it. Or I mi ght ascri be a pr oper t y t o somet hi ng else, or even t o mysel f ,
under a r el at i on of acquai nt ance I bear t o t hat thing.13 Surel y t he pr oper t i es
that give t he cont ent of at t i t udes in t hese ways c a nnot be rel i ed on t o be
perfect l y nat ur al , hence cannot be r epl aced by uni versal s. I t is i nt erest i ng
to ask whet her t here is any l ower limit t o t hei r nat ural ness (see t he final section
of this paper ) , but surel y no ver y exact i ng s t andar d is possi bl e. Her e agai n
properties are ri ght f or t he j ob, uni versal s are not .
One Over Many. Ar ms t r ong' s mai n ar gument f or uni versal s is t he ' One
over Many' . I t is because I find t hi s ar gument unconvi nci ng t hat I a m
investigating al t er nat i ve r easons t o accept a t heor y of uni versal s.
Her e is a conci se s t at ement of t he ar gument , t a ke n by condens at i on f r om
"Against ' Ost r i ch' Nomi nal i s m", pp. 440-441. A ver y si mi l ar st at ement coul d
have been dr awn f r om t he openi ng pages o f Universals.
I woul d wish t o st art by saying t hat ma ny di f f er ent part i cul ars can all have
what appear s t o be t he s ame nat ur e and dr aw t he concl usi on t hat , as a
result, t her e is a pr i ma f aci e case f or post ul at i ng uni versal s. We ar e
cont i nual l y t al ki ng about di f f er ent t hi ngs havi ng t he s ame pr ope r t y or
qual i t y, bei ng of t he same sor t or ki nd, havi ng t he s ame nat ur e, and so
on. Phi l osopher s dr aw t he di st i nct i on bet ween sameness of t oken and
sameness of t ype. But t hey ar e onl y ma ki ng explicit a di st i nct i on whi ch
ordi nary l anguage (and so, or di nar y t hought ) perfect l y recognises. I suggest
t hat t he f act of sameness of t ype is a Moor e a n fact : one o f t he ma n y fact s
which even phi l osophers shoul d not deny, what ever phi l osophi cal account
or anal ysi s t hey give of such fact s. Any compr ehensi ve phi l os ophy mus t
try t o give some account of Moor ean facts. They const i t ut e t he compul sor y
quest i ons in t he phi l osophi cal exami nat i on paper .
From this poi nt of depar t ur e, Ar ms t r ong makes his case by criticising ri val
attempts t o answer t he compul s or y quest i on, and by rej ect i ng views t hat
decline t o answer it at all.
Still mor e conci sel y, t he One over Ma n y pr obl e m is pr esent ed as t he
problem of gi vi ng s ome account of Mo o r e a n fact s o f appar ent sameness of
type. Thus under s t ood, I agree t hat t he quest i on is compul s or y; I agree t hat
1t See my 'Attitudes De Dicto and De Se', Philosophical Review 88 (1979) pp. 513-543; and
'Individuation by Acquaintance and by Stipulation'.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
352 New Work f or a Theory of Universals
Ar mst r ong' s post ul at i on of shared uni versal s answers it; but I t hi nk t hat an
adequat e Nomi nal i sm also answers it.
An ef f or t at syst emat i c phi l osophy must i ndeed give an account of any
pur por t ed fact . Ther e are t hr ee ways t o give an account . (1) ' I deny it' -
this earns a failing mar k i f t he fact is real l y Moor ean. (2) ' I anal yse it thus'
- - this is Ar mst r ong' s response t o t he fact s of appar ent sameness of type.
Or (3) ' I accept it as pri mi t i ve' . Not ever y account is an analysis! A system
t hat t akes cert ai n Moor e a n fact s as pri mi t i ve, as unanal ysed, cannot be
accused of failing t o make a pl ace f or t hem. It nei t her shirks t he compul sory
quest i on nor answers i t by deni al . It does give an account .
An adequat e Nomi nal i sm, of course, is a t heor y t hat t akes Moor e a n facts
of appar ent sameness of t ype as pri mi t i ve. It predi cat es mut ual resemblance
of t he t hi ngs whi ch are appar ent l y of t he same t ype; or it predicates
nat ural ness of some pr oper t y t hat t hey all share, i.e. t hat has t hem all as
member s; and it declines t o anal yse t hese pr edi cat i ons any f ur t her . That is
why t he pr obl em of One over Many, ri ght l y under s t ood, does not provide
mor e t han a pri ma facie r eason t o post ul at e uni versal s. Uni versal s afford
one sol ut i on, but t her e are ot hers.
I f ear t hat t he pr obl em does not r emai n ri ght l y under s t ood. Ear l y in
Universals it undergoes an unf or t unat e doubl e t r ansf or mat i on. In t he course
of a few pages (Universals, I, pp. 11-16) t he legitimate demand f or an account
of Moor e a n fact s of appar ent sameness of t ype t ur ns i nt o a demand f or an
analysis of pr edi cat i on in general. The analysandum becomes t he schema
' a has t he pr oper t y F' . The t ur ni ng poi nt t akes onl y t wo sentences (p. 12):
How is [the Nomi nal i st ] t o account f or t he appar ent (i f usual l y partial)
i dent i t y of numeri cal l y di fferent particulars? How can t wo di fferent things
bot h be whi t e or bot h be on a t abl e?
And ver y soon (pp. 16-17) t hose who ' refuse t o count enance universals but
who at t he same t i me see no need f or any reduct i ve analyses [of t he schema
of pr edi cat i on] ' , t hose accor di ng t o whom ' t here are no universals but the
pr opos i t i on t hat a is F i s per f ect l y all ri ght as it is' st and accused of dodging
t he compul s or y quest i on.
When t he demand f or an account - f or a pl ace in one' s syst em - turned
i nt o a demand f or an analysis, t hen I say t hat t he quest i on ceased t o be
compul sor y. And when t he analysandum swi t ched, f r om Moor ean facts of
appar ent sameness of t ype t o pr edi cat i on general l y, t hen I say t hat the
quest i on ceased t o be answerabl e at all. The t r ans f or med pr obl em of One
over Ma ny deserves our neglect. The ost ri ch t hat will not l ook at it is a wise
bi r d i ndeed.
Despi t e his words, I do not t hi nk t hat Ar mst r ong really means t o demand,
ei t her f r om Nomi nal i st s or f r om hi msel f, a f ul l y general analysis of
pr edi cat i on. For none is so r eady as he t o insist t hat not j ust any shared
pr edi cat e makes f or even appar ent sameness of t ype. ( That is what gives his
t heor y its distinctive i nt erest and meri t . ) It woul d be bet t er t o put the
t r ans f or med pr obl em t hus: one way or anot her , all pr edi cat i on is to be
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 353
analysed. Some predi cat i ons are t o be anal ysed away in t er ms of ot hers. Her e
we have one - of f anal yses f or specific pr edi cat es - as it mi ght be, f or ' gr ue' .
But all t hose predi cat i ons t hat r emai n, af t er t he one- of f anal yses are finished,
are t o be anal ysed whol esal e by means o f a general anal ysi s o f t he s chema
' a has pr oper t y F' .
Ther e is t o be no unanal ysed pr edi cat i on. Ti me and agai n, Ar ms t r ong
wields t hi s r equi r ement agai nst ri val t heori es. One t heor y af t er anot her falls
victim t o t he ' r el at i on regress' : in t he cour se of anal ysi ng ot her pr edi cat i ons,
the t heor y has r esor t t o a new pr edi cat e t hat cannot , on pai n of ci rcul ari t y,
be anal ysed al ong wi t h t he rest. So falls Cl ass Nomi nal i s m (i ncl udi ng t he
version wi t h pr i mi t i ve nat ur al ness t hat I deem adequat e) : it empl oys
predi cat i ons o f class member s hi p, whi ch pr edi cat i ons it cannot wi t hout
circularity anal yse in t er ms of class member s hi p. So falls Res embl ance
Nomi nal i sm: it fails t o anal yse pr edi cat i ons o f r esembl ance. So fal l var i ous
other, less deservi ng Nomi nal i s ms . And so fal l r i val f or ms of Real i sm, f or
instance Tr ans cendent , Pl at oni c Real i sm: t hi s t i me, pr edi cat i ons of
part i ci pat i on evade anal ysi s. Specific t heori es meet ot her , specific obj ect i ons;
suffice it t o say t hat I t hi nk t hese i nconcl usi ve agai nst t he t wo Nomi nal i s ms
that I cal l ed adequat e. But t he cl i ncher, t he one ar gument t hat recurs
t hr oughout t he ma ny r ef ut at i ons, is t he r el at i on regress. And t hi s a mount s
to the obj ect i on t hat t he t heor y under at t ack does not achi eve its pr es umed
aim of doi ng away wi t h all unanal ysed pr edi cat i on and t her ef or e fails t o solve
the t r a ns f or me d pr obl e m of One over Many.
Doi ng away wi t h all unanal ys ed pr edi cat i on is an unat t ai nabl e ai m, and
so an unr eas onabl e ai m. No t heor y is t o be f aul t ed f or fai l i ng t o achi eve it.
For how coul d t here be a t heor y t hat names entities, or quant i fi es over t hem,
in the cour se of its sentences, and yet al t oget her avoi ds pr i mi t i ve predi cat i on?
Artificial t ri cks asi de, 14 t he t hi ng cannot be done.
What ' s t r ue is t hat a t heor y ma y be f aul t ed f or its ove r a bunda nt pr i mi t i ve
predications, or f or undul y myst eri ous ones, or f or undul y compl i cat ed ones.
These are not f at al faul t s, however . The y ar e t o be count ed agai nst a t heor y,
along wi t h its f aul t s of over l y gener ous ont ol ogy or o f di sagr eement wi t h
l ess-t han-Moorean commons ens i cal opi ni ons. Ri val phi l osophi cal t heori es
have t hei r pri ces, whi ch we seek t o meas ur e. But i t ' s all t oo cl ear t hat f or
phi l osophers, at l east , t her e ai n' t no such t hi ng as a free l unch.
How does Ar ms t r ong hi msel f do wi t hout pr i mi t i ve pr edi cat i on? - He
doesn' t. Consi der t he pr edi cat e ' i nst ant i at es' (or ' has' ) , as in ' par t i cul ar a
14 Let S be the syntactic category of sentences, let N be the category of names, and for any
categories x and y, let x/y be the category of expressions that attach to y-expressions to
make x-expressions. Predicates, then, are category S/N. (Or (S/N)/Nfor two-place predicates,
and so on.) To embed names (or variables in the category of names) into sentences without
primitive predication, take any category q which is neither S nor N, nor S/N, and let there
be primitives of categories Q/N and S/Q. Or take Q~ and Qz, different from S and N and
S/N and each other, and let the primitives be of categories Q1/N, Qz/Q~, and S/Q2. Or
. . . . I cannot see how this trickery could be a genuine alternative to, rather than a disguise
for, primitive predication.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
354 New Work f or a Theory of Universals
instantiates universal F' or ' this el ect ron has unit charge' . No one- of f analysis
applies t o this specific predi cat e. ' Such i dent i t y in nat ur e [as results f r om the
havi ng of one uni versal in many part i cul ars] is l i t eral l y i nexpl i cabl e, in the
sense t hat it cannot be f ur t her expl ai ned. ' (Universals, I, p. 109.) Neither
do pr edi cat i ons of ' i nst ant i at es' fall under Ar mst r ong' s general analysis of
(ot herwi se unanal ysed) pr edi cat i on. Hi s is a non-rel at i onal Realism: he
declines, with good reason, t o post ul at e a dyadi c uni versal of instantiation
t o bi nd part i cul ars t o t hei r universals. (And i f he did, it woul d onl y postpone
t he need f or pri mi t i ve pr edi cat i on. ) So l et all who have fel t t he bite of
Ar mst r ong' s r el at i on regress rise up and cr y ' Tu quoque! ' And let us mark
well t hat Ar ms t r ong is pr epar ed t o give one pr edi cat e ' what has been said
t o be t he privilege of t he har l ot : power wi t hout responsi bi l i t y. The predicate
is i nf or mat i ve, it makes a vital cont r i but i on t o telling us what is t he case,
t he wor l d is di f f er ent i f it is di f f er ent , yet ont ol ogi cal l y it is supposed not
t o commi t us. Ni ce work: i f you can get it. ' ( Compar e Ar ms t r ong on Quine's
t r eat ment of pr edi cat i on, "Agai nst ' Ost ri ch' Nomi nal i sm", p. 443.)
Let us dump t he pr oj ect of getting rid of pri mi t i ve pr edi cat i on, and return
t o t he sensible - t hough not compul s or y - pr oj ect of anal ysi ng Moorean
fact s of appar ent sameness of t ype. Now does t he r el at i on regress serve
Ar ms t r ong bet t er ? I t hi nk not . It does make bet t er sense wi t hi n t he more
sensible pr oj ect , but it still bites Ar ms t r ong and his rivals wi t h equal force.
Let t he Nomi nal i st say ' These donkeys resembl e each ot her , so likewise do
t hose stars, and t her e analysis ends. ' Let t he Pl at oni st say ' Thi s statue
part i ci pat es in t he For m of beaut y, likewise t hat l ect ure part i ci pat es in the
For m of t r ut h, and t her e analysis ends. ' Le t Ar ms t r ong say ' Thi s electron
i nst ant i at es uni t charge, likewise t hat pr ot on i nst ant i at es t ri part i t eness, and
t her e analysis ends. ' It is possible t o compl ai n in each case t hat a fact of
sameness of t ype has gone unanal ysed, t he t ypes bei ng respectively
resembl ance, par t i ci pat i on, and i nst ant i at i on. But it is f ar f r om evi dent that
t he alleged facts are Moor ean, and still less evi dent t hat t he first t wo are more
Moor ean t han t he t hi rd. None of t hem are r emot el y t he equals of t he genuine
Moor ean f act t hat , in some sense, di f f er ent l umps of gol d are t he same in
ki nd.
Mi chael Devi t t has denounced t he One over Ma ny pr obl em as a mirage
bet t er l eft unseen. 15 I have f ound Devitt' s discussion i nst ruct i ve and I agree
wi t h much of what he says. But Devi t t has j oi ned Ar ms t r ong in transforming
t he One over Many pr obl em. He t akes it t o be t he pr obl em of analysing the
schema
a and b have t he same pr oper t y (are of t he same t ype), F-ness
ot herwi se t han by means o f a one - of f analysis f or some specific F. To that
pr obl em it is fai r t o answer as he does t hat
" 'Ostrich Nominalism' or 'Mirage Realism'?", Pacific Philosophical Quarterly 61 (1980)
pp. 433-439. Devitt speaks on behalf of Quine as well as himself; Quine indicates agreement
with Devitt in 'Soft Impeachment Disowned'.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis
a is F; b is F
355
is analysis enough, once we give over t he ai m of doi ng wi t hout pri mi t i ve
predication. But Devi t t has set hi msel f t oo easy a pr obl em. I f we at t end t o
the modest , unt r ans f or med One over Many pr obl em, whi ch is no mi rage,
we will ask about a di f f er ent analysandum:
a and b have some c ommon pr oper t y (are s omehow of t he same t ype)
in which it is not said what a and b have in common. Thi s less definite
analysandum is not cover ed by what Devi t t has said. I f we t ake a cl earl y
Moorean case, he owes us an account : ei t her an analysis or an over t resort
to primitive pr edi cat i on of resembl ance.
Duplication, Supervenience, and Divergent Worlds. Hencef or t h I shall
speak onl y of my need f or t he di st i nct i on bet ween nat ur al and unnat ur al ,
or more and less nat ur al , pr oper t i es. It is t o be under s t ood t hat t he wor k
I have in st ore f or an adequat el y di scr i mi nat or y t heor y of propert i es mi ght
be new wor k f or a t heor y of uni versal s, or it mi ght i nst ead be wor k f or t he
resources of an adequat e Nomi nal i sm.
I begin with t he pr obl em of anal ysi ng dupl i cat i on. We are fami l i ar with
cases of appr oxi mat e dupl i cat i on, e.g. when we use copyi ng machi nes. And
we under st and t hat i f these machi nes were mor e per f ect t han t hey are, t he
copies t hey made woul d be per f ect dupl i cat es of t he original. Copy and
original woul d be alike in size and shape and chemi cal composi t i on of t he
ink marks and t he paper , alike in t emper at ur e and magnet i c al i gnment and
electrostatic charge, alike even in t he exact ar r angement of t hei r el ect rons
and quarks. Such duplicates woul d be exactly alike, we say. They woul d mat ch
perfectly, t hey woul d be qual i t at i vel y identical, t hey woul d be indiscernible.
But t hey woul d not have exact l y t he same pr oper t i es, in my sense of t he
word. As in t he case of any t wo things, count l ess class boundar i es woul d
divide t hem. Int ri nsi cal l y, leaving out t hei r rel at i ons t o t he rest of t he worl d,
they woul d be j ust alike. But t hey woul d occupy di f f er ent spat i o- t empor al
positions; and t hey might have di fferent owners, be first examined in di fferent
centuries, and so on.
So i f we wish t o anal yse dupl i cat i on in t er ms of shared propert i es, it seems
that we must first distinguish t he intrinsic (or ' i nt ernal ' ) propert i es f r om t he
extrinsic (or ' ext ernal ' or ' rel at i onal ' ) propert i es. Then we may say t hat t wo
things are dupl i cat es i f f t hey have precisely t he same intrinsic propert i es,
however much t hei r extrinsic pr oper t i es mi ght di f f er . But our new pr obl em
of dividing t he pr oper t i es i nt o intrinsic and extrinsic is no easier t han our
original pr obl em of anal ysi ng dupl i cat i on. In fact , t he t wo pr obl ems are
joined in a t i ght little circle of i nt erdefi nabi l i t y. Dupl i cat i on is a mat t er of
sharing intrinsic propert i es; intrinsic propert i es are j ust t hose propert i es t hat
never di ffer bet ween dupl i cat es. Pr ope r t y P is intrinsic i f f, f or any t wo
duplicate things, not necessarily f r om t he same worl d, ei t her bot h have P
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
356 Ne w Wo r k f o r a Theor y o f Universals
o r n e i t h e r d o e s . P i s e x t r i n s i c i f f t h e r e i s s o me s u c h p a i r o f d u p l i c a t e s o f whi c h
o n e h a s P a n d t h e o t h e r l a c k s p . 1 6
I f we r e l i e d o n o u r p h y s i c a l t h e o r y t o b e a c c u r a t e a n d e x h a u s t i v e , we mi g h t
t h i n k t o d e f i n e d u p l i c a t i o n i n p h y s i c a l t e r ms . We b e l i e v e t h a t d u p l i c a t e s mu s t
b e a l i k e i n t h e a r r a n g e me n t o f t h e i r e l e c t r o n s a n d q u a r k s - - wh y n o t put
t h i s f o r wa r d a s a d e f i n i t i o n ? B u t s u c h a ' d e f i n i t i o n ' i s n o a n a l y s i s . It
p r e s u p p o s e s t h e p h y s i c s o f o u r a c t u a l wo r l d ; h o we v e r p h y s i c s i s c o n t i n g e n t
a n d k n o wn a p o s t e r i o r i . T h e d e f i n i t i o n d o e s n o t a p p l y t o d u p l i c a t i o n at
p o s s i b l e wo r l d s wh e r e p h y s i c s i s d i f f e r e n t , o r t o d u p l i c a t i o n b e t we e n wo r l d s
t h a t d i f f e r i n t h e i r p h y s i c s . No r d o e s i t c a p t u r e wh a t t h o s e i g n o r a n t o f phys i cs
me a n wh e n t h e y s p e a k - a s t h e y d o - o f d u p l i c a t i o n .
T h e p r o p e r c o u r s e , I s u g g e s t , i s t o a n a l y s e d u p l i c a t i o n i n t e r ms o f s h a r e d
p r o p e r t i e s ; b u t t o b e g i n n o t wi t h t h e i n t r i n s i c p r o p e r t i e s b u t r a t h e r wi t h n a t u r a l
p r o p e r t i e s . T wo t h i n g s a r e q u a l i t a t i v e d u p l i c a t e s i f t h e y h a v e e x a c t l y t h e s a me
p e r f e c t l y n a t u r a l p r o p e r t i e s . 17
P h y s i c s i s r e l e v a n t b e c a u s e i t a s p i r e s t o g i v e a n i n v e n t o r y o f n a t u r a l
16 Given duplication, we can also subdivide t he extrinsic properties, distinguishing pure cases
from various mixtures of extrinsic and intrinsic. Part i t i on the things, of this and other worlds,
i nt o equivalence classes under t he rel at i on of duplication. A propert y may divide an
equivalence class, may include it, or may exclude it. A propert y P is extrinsic, as we said,
if it divides at least some of t he classes. We have four subcases. (1) P divides every class;
t hen we may call Ppurel y extrinsic. (2) P divides some classes, includes some, and excludes
none; t hen P is t he di sj unct i on of an intrinsic propert y and a purely extrinsic property. (3)
P divides some, excludes some, and includes none; t hen P is t he conj unct i on of an intrinsic
propert y and a purely extrinsic propert y. (4) P divides some, includes some, and excludes
some; t hen P is t he conj unct i on of an intrinsic propert y and an impurely extrinsic property
of t he sort considered in t he second case, or equivalently is t he di sj unct i on of an intrinsic
propert y and an impurely extrinsic propert y of t he sort considered in t he t hi rd case.
We can also classify relations as intrinsic or extrinsic, but in t wo di fferent ways. Take
a dyadic relation, i.e. a class or ordered pairs. Call t he rel at i on intrinsic to its relata iff,
whenever a and a' are duplicates (or identical) and b and b ' are duplicates (or identical),
then bot h or neither of the pairs < a, b > and < a ' , b ' > stand in the relation. Call the relation
intrinsic" to itspairs iff, whenever the pairs <a, b > and <a' , b" > themselves are duplicates,
then bot h or neither of t hem st and in t he relation. In t he second case, a stronger requirement
is imposed on the pairs. For instance they might fail to be duplicate pairs because the distance
between a and b differs from t he distance between a ' and b ' , even t hough a and a' are
duplicates and b and b ' are duplicates. In t radi t i onal t ermi nol ogy, ' i nt ernal relations' are
intrinsic t o t hei r relata; ' external rel at i ons' are intrinsic t o t hei r pairs but not t o t hei r relata;
and relations extrinsic even t o their pairs, such as the relation of belonging to the same owner,
get left out of t he classification altogether.
Our definition of intrinsic properties in terms of duplication closely resembles the definition
of ' differential properties' given by Michael Slote in ' Some Thought s on Goodman' s Riddle',
Analysis 27 (1967) pp. 128-132, and in Reason and Scepticism (George Allen & Unwin, 1970).
But where I quantify over possibilia, Slote applies modality to ordinary, presumably actualist,
quantifiers. That makes a difference. An extrinsic propert y mi ght differ between duplicates,
but only when t he duplicates i nhabi t di fferent worlds; t hen Slote would count t he property
as differential. An example is t he propert y of being a sphere t hat i nhabi t s a world where
t here are pigs or a cube t hat i nhabi t s a world wi t hout pigs.
See my ' Extrinsic Propert i es' , Philosophical Studies 44 (1983) for f ur t her discussion of
t he circle from dupl i cat i on t o intrinsicness and back.
17 Likewise <a, b> and <a' , b" > and duplicate pairs i ff a and a' have exactly t he same
perfectly nat ur al properties, and so do b and b' , and also t he perfectly nat ural relations
between a and b are exactly the same as t hose between a ' and b ' .
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewi s 357
properties -- not a compl et e i nvent or y, per haps, but a compl et e enough
inventory t o account f or dupl i cat i on among actual things. I f physics succeeds
in this, t hen dupl i cat i on wi t hi n our wor l d amount s t o sameness of physical
description. But t he nat ur al pr oper t i es t hemsel ves are what mat t er , not t he
theory t hat tells us what t hey are. I f Mat eri al i sm were false and physics an
utter fai l ure, as is t he case at some depl or abl e worl ds, t her e woul d still be
duplication in vi rt ue of shared nat ur al pr oper t i es.
On my analysis, all per f ect l y nat ur al propert i es come out intrinsic. That
seems ri ght . The converse is not t rue. Int ri nsi c pr oper t i es may be di sj unct i ve
and mi scel l aneous and unnat ur al , so l ong as t hey never di f f er bet ween
duplicates. The perfect l y nat ur al propert i es compri se a basis f or t he intrinsic
properties; but ar bi t r ar y Bool ean compounds of t hem, however unnat ur al ,
are still intrinsic. Hence i f we adopt t he sort of adequat e Nomi nal i sm t hat
draws a pri mi t i ve di st i nct i on bet ween nat ur al and unnat ur al propert i es, t hat
is not t he same t hi ng as drawi ng a pri mi t i ve di st i nct i on bet ween intrinsic and
extrinsic properties. The f or mer distinction yields t he latter, but not vi ce versa.
Likewise i f we adopt t he sort of adequat e Nomi nal i sm t hat begins wi t h
a suitable rel at i on of part i al resembl ance, t hat is not t he same t hi ng as t aki ng
duplication i t sel f as pri mi t i ve. Agai n, t he f or mer yields t he l at t er, but not
vice versa.
If i nst ead we rej ect Nomi nal i sm, and we t ake t he per f ect l y nat ur al
properties t o be t hose t hat cor r espond t o universals (in t he sense t hat t he
members of t he pr oper t y are exact l y t hose things t hat i nst ant i at e t he
universal), t hen all t he propert i es t hat cor r espond t o universals are intrinsic.
So are all t he Bool ean compounds -- di sj unct i ons, negat i ons, etc. -- of
properties t hat cor r espond t o universals. The universals t hemsel ves are
intrinsic e x of f i ci o, so t o speak.
But her e I must confess t hat t he t heor y of universals f or whi ch I of f er new
work cannot be exact l y Ar mst r ong' s t heor y. For it must rej ect extrinsic
universals; whereas Ar mst r ong admi t s t hem, al t hough not as irreducible. (See
Uni versal s, II, pp. 78-79.) I t hi nk he woul d be bet t er of f wi t hout t hem, given
his own aims. (1) They subvert t he desi red connect i on bet ween shari ng of
universals and Moor ean facts of part i al or t ot al sameness of nat ur e.
Admittedly, t her e is such a t hi ng as resembl ance in ext ri nsi c respects: things
can be alike in t he roles t hey pl ay vi s - a- vi s ot her t hi ngs, or in t he origins
they spring f r om. But such resembl ances are not what we mean when we
say of t wo things t hat t hey are of t he same ki nd, or have t he same nat ur e.
(2) They subvert t he desired i mmanence of universals: i f somet hi ng
instantiates an extrinsic uni versal , t hat is not a fact j ust about t hat thing.
(3) They are not needed f or Ar mst r ong' s t heor y of laws of nat ur e; any
supposed law connect i ng ext ri nsi c universals of things can be equi val ent l y
replaced by a l aw connect i ng intrinsic st ruct ures of l arger systems t hat have
those t hi ngs as part s.
Thus I am cont ent t o say t hat i f t her e are uni versal s, intrinsic dupl i cat es
are things havi ng exact l y t he same universals. We need not say ' . . . exact l y
the same i nt r i ns i c universals' because we shoul d not believe in any ot her kind.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
358 Ne w Wor k f o r a Theory o f Universals
Not onl y is dupl i cat i on of i nt erest in its own right; it also is needed in
deal i ng wi t h ot her t opi cs in met aphysi cs. Hence such t opi cs cr eat e a derived
need f or nat ur al propert i es. I shall consi der t wo t opi cs where I find need to
speak of dupl i cat i on: superveni ence and di vergent worl ds.
First, superveni ence. A superveni ence thesis is a deni al of i ndependent
var i at i on. Gi ven an ont ol ogy of pos s i bi l i a, we can f or mul at e such theses in
t er ms of di fferences bet ween possible i ndi vi dual s or worl ds. To say t hat so-
and-so supervenes on such-and-such is t o say t hat t her e can be no difference
in respect of so-and-so wi t hout di fference in respect of such-and-such. Beauty
of st at ues supervenes on t hei r shape, size, and col our , f or i nst ance, i f no
t wo st at ures, in t he same or di f f er ent worl ds, ever di f f er in beaut y without
also di f f er i ng in shape or size or col our . TM
A superveni ence thesis is, in a br oad sense, reductionist. But it is a stripped-
down f or m or r educt i oni sm, unencumber ed by dubi ous denials of existence,
claims of ont ol ogi cal pr i or i t y, or claims of t ransl at i bi l i t y. One mi ght wish
t o say t hat in some sense t he beaut y of st at ues is not hi ng over and above
t he shape and size and col our t hat behol ders appreci at e, but wi t hout denying
t hat t her e is such a t hi ng as beaut y, wi t hout cl ai mi ng t hat beaut y exists only
in some l ess-t han-fundament al way, and wi t hout under t aki ng t o paraphrase
ascri pt i ons of beaut y in t er ms of shape etc. A superveni ence thesis seems
t o capt ur e what t he caut i ous r educt i oni st wishes t o say.
Even i f reduct i oni st s ought t o be less caut i ous and ai m f or translation,
still it is a good i dea t o at t end t o t he quest i on of superveni ence. For if
superveni ence fails, t hen no scheme of t r ansl at i on can be cor r ect and we
needn' t go on Chi shol mi ng away in search of one. I f superveni ence succeeds,
on t he ot her hand, t hen some cor r ect scheme must exist; t he remaining
quest i on is whet her t her e exists a cor r ect scheme t hat is less t han infinitely
compl ex. I f beaut y is superveni ent on shape et c. , t he wor st t hat can happen
is t hat an ascri pt i on of beaut y is equi val ent t o an uncount abl y infinite
di sj unct i on of maxi mal l y specific descriptions of shape etc., which descriptions
mi ght t hemsel ves i nvol ve infinite conj unct i ons.
Int erest i ng superveni ence theses usual l y i nvol ve t he not i on of qualitative
dupl i cat i on t hat we have j ust consi dered. Thus we ma y ask what does or
doesn' t supervene on t he qualitative charact er of t he entire worl d, t hroughout
all of hi st or y. Suppose t hat t wo possible worl ds are per f ect qualitative
dupl i cat es - must t hey t hen also have exact l y t he same di st ri but i ons of
obj ect i ve probabi l i t y, t he same laws of nat ur e, t he same count er f act ual s and
causal relations? Must t hei r inhabitants have t he same de re modal properties?
I f so, it makes sense t o pur sue such pr oj ect s as a f r equency analysis of
probability, a regularity analysis of laws of nat ure, or a comparat i ve similarity
~s For a general discussion of supervenience, see Jaegwon Kim, 'Supervenience and Nomological
Incommensurables', American Philosophical Quarterly 15 (1978) pp. 149-156.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 359
analysis of causal count er f act ual s and de re modal i t y. I f not , such pr oj ect s
are doomed f r om t he st art , and we needn' t l ook at t he details of t he at t empt s.
But we cannot even raise these quest i ons of superveni ence unless we can speak
of dupl i cat e worl ds. And t o do t hat , I have suggest ed, we need nat ur al
properties.
(Not e t hat i f possi bl e worl ds obey a pr i nci pl e o f i dent i t y of qual i t at i ve
indiscernibles, t hen all t hese super veni ence t heses hol d aut omat i cal l y. I f no
two wor l ds ar e dupl i cat es, t hen af ort i ori no t wo are dupl i cat es t hat di f f er
in t hei r pr obabi l i t i es, l aws . . . . . or anyt hi ng else.)
We mi ght al so ask whet her qual i t at i ve char act er super venes on anyt hi ng
less. For i nst ance, we mi ght ask whet her gl obal qual i t at i ve char act er
supervenes on l ocal qual i t at i ve char act er . Say t hat t wo worl ds ar e local
duplicates i f f t hey are divisible i nt o cor r es pondi ng smal l par t s in such a way
that (1) cor r es pondi ng par t s of t he t wo wor l ds are dupl i cat es, and (2) t he
correspondence pr eser ves s pat i ot empor al rel at i ons. (The exact meani ng
depends, of course, on what we mean by ' smal l ' . ) I f t wo wor l ds are l ocal
duplicates, t hen mus t t hey be dupl i cat es simpliciter? Or coul d t hey di f f er in
ways t hat do not pr event l ocal dupl i cat i on - e.g. in ext ernal rel at i ons, ot her
than t he s pat i ot empor al r el at i ons t hemsel ves, bet ween separ at ed t hi ngs?
Again, we mus t ma ke sense of dupl i cat i on - t hi s t i me, bot h in t he l arge
and in t he smal l - even t o ask t he quest i on. 19
Next, di vergent worl ds. I shall say t hat t wo possi bl e worl ds diverge i f f t hey
are not dupl i cat es but t hey do have dupl i cat e initial t empor al segment s. Thus
our wor l d and anot her mi ght ma t c h per f ect l y up t hr ough t he year 1945, and
go t hei r separ at e ways t her eaf t er .
Not e t hat we need no i dent i t y of t i mes across worl ds. Our wor l d t hr ough
our 1945 dupl i cat es an initial segment o f t he ot her worl d; t hat ot her wor l dl y
segment ends wi t h a year t hat i ndeed r esembl es our 1945, but it is par t of
otherworldly t i me, not par t of our t i me. Al so, we need no separ at i on of t i me
and space t hat cont r avenes Rel at i vi t y - we have initial t e mpor a l segment s,
of this or anot her wor l d, i f we have s pat i ot empor al regi ons bounde d by
spacelike surfaces t hat cut t he wor l d in t wo.
I distinguish divergence of worl ds f r om branching of worl ds. I n br anchi ng,
instead of dupl i cat e segment s, one and t he s ame initial segment is al l egedl y
shared as a c o mmo n par t by t wo over l appi ng worl ds. Br anchi ng is
probl emat i c in ways t hat di ver gence is not . Fi rst , because an i nhabi t ant of
the shared segment cannot speak unequi vocal l y of the worl d he lives in. What
if he says t her e will be a sea fight t omor r ow, meani ng of cour se t o speak
of the f ut ur e of his own wor l d, and one o f t he t wo wor l ds he lives in has
19 Such a thesis of supervenience of the global on the local resembles the 'holographic hypothesis'
considered and rejected by Saul Kripke in 'Identity Through Time', presented at the 1979
conference of the American Philosophical Association, Eastern Division, and elsewhere.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
360
Ne w Wor k f o r a Theory o f Universals
a sea fight the next day and the other doesn' t? Second, because overlap of
worlds interferes with the most salient principle of demarcat i on for worlds,
vi z. that two possible individuals are part of the same world iff they are linked
by some chain of external relations, e. g. of spatiotemporal relations. (I know
of no other example.) Neither of these difficulties seems insuperable, but both
are better avoided. That makes it reasonable to prefer a theory of
nonoverlapping divergent worlds to a t heory of branching worlds. Then we
need to be able to speak of qualitative duplication of world-segments, which
we can do in terms of shared natural properties.
Divergent (or branching) worlds are of use in defining Determinism. The
usual definitions are not very satisfactory. If we say that every event has a
cause, we overlook probabilistic causation under Indeterminism. If we speak
of what could be predicted by a superhuman calculator with unlimited
knowledge of history and the laws of nature, we overlook obstacles that might
prevent prediction even under Determinism, or else we try to make
nonvacuous sense of counterfactuals about what our predictor could do if
he had some quite impossible combi nat i on of powers and limitations.
A better approach is as follows. First, a system of laws of nature is
Deterministic iff no two divergent worlds bot h conform perfectly to the laws
of that system. Second, a world is Deterministic iff its laws comprise a
Deterministic system. Third, Determinism is the thesis that our world is
Deterministic. E
(Alternative versions of Determinism can be defined in similar fashion.
For instance, we could strengthen the first step by prohibiting convergence
as well as divergence of law-abiding worlds. Or we could even require that
no two law-abiding worlds have duplicate moment ary slices without being
duplicates t hroughout their histories. Or we could define a weaker sort of
Determinism: we could call a world f o r t u i t o u s l y Deterministic, even if its
laws do not comprise a Deterministic system, iff no world both diverges from
it and conforms to its laws. The laws and early history of such a world suffice
to determine later history, but only because the situations in which the laws
fall short of Determinism never arise. We might equivalently defirae fortuitous
Determinism as follows: for any historical fact F and any initial segment S
of the world, there are a true proposition H about the history of S-and a
true proposition L about the laws of nature, such t hat H and L together
21
strictly imply F. Does this definition bypass our need to speak of duplication
of initial segments? Not so, for we must ask what it means to say that H
20 This approach is due, in essence, to Richard Montague, 'Deterministic Theories', in Decisions,
Values and Groups, II (Pergamon Press, 1962), and in his Formal Philosophy (Yale University
Press, 1974). But Montague did not speak as I have done of duplication of initial segments
of worlds in virtue of the sharing of certain 61ite properties. Instead, he used sameness of
description in a certain vocabulary, which vocabulary was left as an unspecified parameter
of his analysis. For he wrote as a logician obliged to remain neutral on questions of
metaphysics.
21 A closely related definition appears in Peter van Inwagen, 'The Incompatibility of Free Will
and Determinism', Philosophical Studies 27 (1975) pp. 185-199.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 361
is about t he hi st or y of S. I t ake t hat t o mean t hat H hol ds at bot h or nei t her
of any t wo worl ds t hat bot h begi n wi t h segment s t hat are dupl i cat es o f S.)
Di ver gent wor l ds are i mpor t a nt al so in connect i on wi t h t he sort of
count erfact ual condi t i onal t hat figures in pat t er ns of causal dependence. Such
count erfact ual s t end t o be t empor al l y as ymmet r i c, and t hi s is what gives rise
to t he a s ymme t r y of causat i on itself. Count er f act ual s o f t hi s sort do not
' backt r ack' : it is not t o be sai d t hat i f t he pr esent were di f f er ent a di f f er ent
past woul d have led up t o it, but r at her t hat i f t he pr esent were di f f er ent ,
the same pas t woul d have had a di f f er ent out come. Gi ven a hypot hesi sed
difference at a cer t ai n t i me, t he event s of f ut ur e t i mes nor mal l y woul d be
very di f f er ent i ndeed, but t he event s of past t i mes (except per haps f or t he
very near past ) woul d be no di f f er ent . Thus act ual i t y and its count er f act ual
alternatives are di vergent worl ds, wi t h dupl i cat e initial segment s, z2
Mi ni mal Mat eri al i sm. Ther e is a di ffi cul t y t hat arises i f we at t empt t o
f or mul at e cer t ai n r educt i oni st vi ews, f or i nst ance Mat er i al i s m, as
supervenience theses. A sol ut i on t o this difficulty empl oys nat ur al pr oper t i es
not onl y by way of dupl i cat i on but in a mor e di rect way al so.
Roughl y speaki ng, Mat er i al i sm is t he t hesi s t hat physi cs - somet hi ng not
too di f f er ent f r o m pr es ent - day physi cs, t hough pr es umabl y s omewhat
i mproved - is a compr ehens i ve t heor y of t he wor l d, compl et e as well as
correct. The wor l d is as physi cs says it is, and t her e' s no mor e t o say. Wor l d
history wr i t t en in physi cal l anguage is all of wor l d hi st ory. That is r ough
speaking i ndeed; our goal will be t o give a bet t er f or mul at i on. But bef or e
I try t o say mor e preci sel y what Mat er i al i sm is, let me say what it is not .
(1) Mat er i al i sm is not a t hesi s of finite t r ansl at i bi l i t y of all our l anguage i nt o
the l anguage of physi cs. (2) Mat er i al i sm is not t o be identified wi t h any one
Materialist t heor y of mi nd. I t is a t hesi s t hat mot i vat es a var i et y of t heori es
of mi nd: versi ons of Behavi our i s m, Funct i onal i s m, t he mi nd- body i dent i t y
theory, even t he t heor y t hat mi nd is all a mi st ake. (3) Mat er i al i sm is not j ust
the t heor y t hat t her e are no t hi ngs except t hose r ecogni sed by physi cs. To
be sure, Mat eri al i st s don' t believe in spirits, or ot her such nonphysi cal things.
But ant i mat eri al i st s ma y not believe in spirits ei t her - t hei r compl ai nt needn' t
be t hat physi cs omi t s s ome of t he t hi ngs t hat t her e are. They ma y compl ai n
instead t hat physi cs over l ooks s ome o f t he ways t her e ar e f or physi cal t hi ngs
to di ffer; f or i nst ance, t hey ma y t hi nk t hat physi cal peopl e coul d di f f er in
what t hei r experi ence is like. (4) Tha t suggest s t hat Mat er i al i s m is, at l east
in part , t he thesis t hat t here are no nat ur al propert i es i nst ant i at ed at our worl d
22 See my 'Counterfactual Dependence and Time's Arrow', Nofts 13 (1979) pp. 455-476; Jonathan
Bennett's review of my Counterfactuals, Canadian Journal of Philosophy 4 (t974) pp. 381-402;
P. B. Downing, 'Subjunctive Conditionals, Time Order, and Causation', Proceedings of
the Aristotelian Society 59 (1959) pp. 125-140; Allan Gibbard and William Harper,
'Counterfactuals and Two Kinds of Expected Utility', in C. A. Hooker, J. T. Leach, and
E. F. McClennen, eds., Foundations and Applications of Decision Theory (Reidel, 1978),
and in W. L. Harper, R. Stalnaker, and G. Pearce, eds., Ifs(Reidel, 1981); and Frank Jackson,
'A Causal Theory of Counterfactuals', Australasian Journal of Philosophy 55 (1977) pp. 3-21.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
362
New Work for a Theory of Universals
except t hose recogni sed by physics. That is bet t er, but I t hi nk still not right.
Coul dn' t t her e be a nat ur al pr oper t y X (in t he nat ur e of t he case, it is hard
t o name an exampl ed which is shared by t he physical brai ns in worl ds like
our s and t he i mmat er i al spirits t hat i nhabi t ot her worl ds? Or by thisworldly
quarks and certain ot herworl dl y particles t hat cannot exist under our physics?
Physi cs coul d qui t e pr oper l y make no ment i on of a nat ur al pr oper t y of this
sort . It is enough t o recogni se t he special case appl i cabl e t o our worl d, X-
cum-physicality, br ai nhood or quar khood as it mi ght be. Then i f by physical
pr oper t i es we mean t hose pr oper t i es t hat are ment i oned in t he l anguage of
physics, a Materialist ought not t o hol d t hat all nat ural propert i es instantiated
in our wor l d are physi cal pr oper t i es.
At this poi nt , it ought t o seem advi sabl e t o f or mul at e Mat eri al i sm as a
superveni ence thesis: no di f f er ence wi t hout physical di fference. Or,
cont r aposi ng: physi cal dupl i cat es are dupl i cat es simpliciter. Afortiori, no
ment al di fference wi t hout physical di fference; physical duplicates are mental
dupl i cat es. The thesis mi ght best be t aken as appl yi ng t o whol e possible
worl ds, in or der t o bypass such quest i ons as whet her ment al life is t o some
ext ent extrinsic t o t he subj ect . So we have this first of several at t empt ed
f or mul at i ons of Mat eri al i sm:
M1. Any t wo possible worl ds t hat are exact l y alike in all respects
recogni sed by physi cs are qual i t at i ve dupl i cat es.
But this will not do. In maki ng Mat eri al i sm i nt o a thesis about how j ust any
t wo worlds can and cannot differ, M1 puts Materialism f or war d as a necessary
t r ut h. That is not what Mat eri al i st s i nt end. Mat eri al i sm is meant t o be a
cont i ngent thesis, a meri t of our wor l d t hat not all ot her worl ds share. Two
worlds coul d i ndeed di f f er wi t hout di fferi ng physically, i f a least one of them
is a wor l d where Mat eri al i sm is false. For i nst ance, our Mat eri al i st i c world
di ffers f r om a nonmat eri al i st i c worl d t hat is physically j ust like ours but that
also cont ai ns physi cal l y epi phenomenal spirits.
Ther e is a noncont i ngent superveni ence thesis near by t hat mi ght appeal
t o Materialists:
M2. Ther e is no di fference, afortiori no ment al di fference, wi t hout some
nonment al difference. Any t wo worlds alike in all nonment al respects
are duplicates, and in part i cul ar do not di ffer in respect of t he mental
lives of t hei r i nhabi t ant s.
Thi s seems t o capt ur e our t hought t hat t he ment al is a pat t er n in a medium,
obt ai ni ng in vi r t ue of l ocal feat ures of t he medi um ( neur on firings) and
per haps also ver y gl obal feat ures (laws of nat ur e) t hat are t oo small or too
big t o be ment al t hemsel ves. But M2 is not Mat eri al i sm. It is bot h less and
mor e. Less, obvi ousl y, because it never says t hat t he medi um is physical.
Mor e, because it denies t he ver y possibility of what I shall call Panpsychistic
Mat eri al i sm.
It is of t en not ed t hat psychophysi cal i dent i t y is a t wo- way street: i f all
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 363
mental pr oper t i es are physi cal , t hen some physi cal propert i es are ment al . But
perhaps not j ust s ome but all physi cal pr oper t i es mi ght be ment al as well;
and i ndeed every pr oper t y of anyt hi ng mi ght be at once physi cal and ment al .
Suppose t her e are i ndeed worl ds where this is so. I f so, pr es umabl y t here
are ma n y such worl ds, not all dupl i cat es, di f f er i ng inter alia in t he ment al
lives of t hei r i nhabi t ant s. But all di f f er ences bet ween such wor l ds are ment al
(as well as physi cal ), so none are nonment al . These wor l ds will be vacuousl y
alike in all nonment al respect s, f or l ack of any nonment al respect s t o di f f er
in. Then M2 fails. And not j ust at t he t r oubl e ma ki ng worl ds; M2 is
noncont i ngent , so i f it fails at any worl ds, it fails at all - even decent
Materialistic worl ds like our s. Maybe Panpsychi st i c Mat er i al i sm is i ndeed
impossible - how do you squar e it wi t h a br oa dl y f unct i onal anal ysi s of
mind? - but a t hesi s t hat says so is mor e t han j ust Mat er i al i sm.
A t hi rd t ry. Thi s much is at l east t rue:
M3. No t wo Mat eri al i st i c worl ds di f f er wi t hout di f f er i ng physi cal l y; any
t wo Mat eri al i st i c wor l ds t hat are exact l y al i ke physi cal l y are
dupl i cat es.
But M3 is not a f or mul at i on of Mat er i al i sm, f or t he di st i nct i on bet ween
Materialistic and ot her worl ds appear s wi t hi n M3. Al l we l earn is t hat t he
Materialistic worl ds compr i s e a class wi t hi n whi ch t her e is no di f f er ence
without physi cal di f f er ence. But t her e ar e ma n y such classes. I n f act any
world, however spi ri t -ri dden, bel ongs t o such a class.
A f our t h t r y. Per haps we shoul d confi ne our at t ent i on t o nomol ogi cal l y
possible worl ds, t hus:
M4. Amo n g worl ds t hat c onf or m t o t he act ual l aws o f nat ur e, no t wo
di f f er wi t hout di f f er i ng physi cal l y; any t wo such wor l ds t hat are
exact l y al i ke physi cal l y are dupl i cat es.
But agai n we have s omet hi ng t hat is bot h less and mor e t han Mat er i al i sm.
Less, because M4 coul d hol d at a wor l d where Mat er i al i sm is false but where
spiritual phenomen~ ar e cor r el at ed wi t h physi cal phe nome na accor di ng t o
strict laws. Mor e, because M4 fails t o hol d at a Materialistic, spi ri t -free wor l d
if the l aws of t hat wor l d do not pr ecl ude t he exi st ence of epi phenomenal
spirits. Our wor l d mi ght be such a wor l d, a wor l d where spirits ar e absent
but not out l awed. 23
So f ar , a super veni ence f or mul at i on of Mat er i al i sm seems elusive. But I
think we can succeed i f we j oi n t he i dea o f super veni ence wi t h t he i dea t hat
a nonmat er i al i st i c wor l d woul d have s omt hi ng ext r a, somet hi ng t hat a
Materialistic worl d lacks. I t mi ght have spirits; or it mi ght have physical things
that di f f er in nonphysi cal ways, f or i nst ance in what t hei r experi ence is like.
2~ This objection against M4 as a formulation of 'the ontological primacy of the microphysical'
appears in Terence Horgan, 'Supervenience and Microphysics', Pacific Philosophical Quarterly
63 (1982) pp. 29-43.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
364 New Work for a Theory of Universals
In ei t her case t her e are ext r a nat ur al pr oper t i es, pr oper t i es i nst ant i at ed in
t he nonmaterialistic worl d but nowhere t o be f ound in t he Materialistic world.
Let us say t hat a pr oper t y is alien t o a wor l d i f f (1) it is not i nst ant i at ed by
any i nhabi t ant of t hat wor l d, and (2) it is not anal ysabl e as a conj unct i on
of , or as a st ruct ural pr oper t y const r uct ed out of , nat ur al pr oper t i es all of
whi ch are i nst ant i at ed by i nhabi t ant s of t hat worl d. (I need t he second clause
because I am fol l owi ng Ar ms t r ong, mutatis mutandis, in decl i ni ng t o rule
out per f ect l y nat ur al propert i es t hat are conj unct i ve or st ruct ural l y complex.
See Universals, II, pp. 30-42 and 67-71. It woul d be wr ong t o count as alien
a compl ex pr oper t y analysable in t erms of nonal i en constituents. ) I f our world
is Mat eri al i st i c, t hen it is safe t o say t hat some of t he nat ur al propert i es
i nst ant i at ed in any nonmat er i al i st i c wor l d are pr oper t i es alien t o our world.
Now we can pr oceed at last t o f or mul at e Mat eri al i sm as a rest ri ct ed and
cont i ngent superveni ence thesis:
M5. Among worl ds where no nat ur al pr oper t i es alien t o our wor l d are
i nst ant i at ed, no t wo di ffer wi t hout di fferi ng physically; any t wo such
worl ds t hat are exact l y alike physi cal l y are dupl i cat es. 24
We t ook Mat eri al i sm t o uphol d t he comprehensi veness of ' somet hi ng not
t oo di f f er ent f r om pr esent - day physics, t hough pr esumabl y somewhat
i mpr oved' . That was del i berat el y vague. Mat eri al i st met aphysi ci ans want to
side wi t h physics, but not t o t ake sides wi t hi n physics. Wi t hi n physics, more
precise claims of compl et eness and correct ness ma y be at issue. Physics
(i gnori ng l at t er - day fai l ures of nerve) is t he science t hat aspires to
comprehensi veness, and par t i cul ar physi cal t heori es may be put f or war d as
fulfilling t hat aspi r at i on. I f so, we must agai n ask what it means t o claim
comprehensi veness. And agai n, t he answer may be gi ven by a superveni ence
f or mul at i on: no di fference wi t hout physical di fference as concei ved by such-
and- such gr and t heor y. But agai n it must be under s t ood as a rest ri ct ed and
cont i ngent superveni ence thesis, appl yi ng onl y among worl ds devoi d of alien
nat ur al pr oper t i es.
Thus t he businesss of physi cs is not j ust t o di scover laws and causal
expl anat i ons. I n put t i ng f or war d as compr ehensi ve t heori es t hat recognise
onl y a l i mi t ed r ange of nat ur al pr oper t i es, physics pr oposes i nvent ori es of
t he nat ur al pr oper t i es i nst ant i at ed in our worl d. Not compl et e inventories,
24 Thi s f or mul at i on r esembl es one pr opos e d by Hor ga n, op. cir. The pr i nci pal di f f er ence is
as f ol l ows. Hor ga n woul d count as al i en ( my t er m, not his) any pr oper t y ci t ed in the
f undament al l aws o f ot her wor l dl y mi cr ophysi cs t hat is not al so expl i ci t l y ci t ed in the
f undament al l aws o f t hi s worl dl y mi crophysi cs. Whet her t he pr oper t y is i nst ant i at ed in either
wor l d does n' t ent er i nt o it. But mus t an al i en pr oper t y figure in l aws o f otherworldly physics?
Mus t it figure i n any ot her wor l dl y laws at all? It seems t hat a Mat eri al i st i c wor l d mi ght differ
wi t hout di f f er i ng physi cal l y f r om a wor l d wher e t her e ar e pr oper t i es al i en i n my sense but
n o t i n Hor ga n' s - - per haps a wor l d wher e l aws ar e i n s hor t suppl y.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 365
perhaps. But compl et e enough t o account f or all t he dupl i cat i ons and
differences t hat coul d ari se in t he absence of al i en nat ur al pr oper t i es. Of
course, t he di scover y o f nat ur al pr oper t i es is i nsepar abl e f r om t he di scover y
of laws. For an excel l ent r eas on t o t hi nk t hat s ome hi t her t o unsuspect ed
nat ural pr oper t i es ar e i nst ant i at ed - pr oper t i es deservi ng of r ecogni t i on by
physics, t he quar k col our s as t hey mi ght be - - is t hat wi t hout t hem, no
sat i sfact ory syst em of l aws can be f ound.
Thi s is r emi ni scent of t he di st i nct i ve a post eri ori , scientific char act er of
Ar mst r ong' s Real i sm (Universals, I, pp. 8-9, and passi m) . But in t he set t i ng
of an ont ol ogy of possibilia, t he di st i nct i on bet ween di scoveri ng what
universals or nat ur al pr oper t i es t here act ual l y are and di scoveri ng whi ch ones
are act ual l y i nst ant i at ed fades away. And t he l at t er quest i on is a post eri ori
on any t heory. What remai ns, and remai ns i mpor t ant , is t hat physics discovers
propert i es. And not j ust any pr oper t i es - nat ur al pr oper t i es. The di scover y
is, f or i nst ance, t hat neut r i nos are not all al i ke. Tha t is not t he di scover y
that di f f er ent ones have di f f er ent pr oper t i es in my sense, bel ong t o di f f er ent
classes. We knew t hat muc h a priori. Rat her , it is t he surpri si ng di scover y
that s ome nat ural pr oper t y di f f er ent i at es s ome neut r i nos f r om ot hers. That
di scovery has in f act been made; I shoul d l i ke t o r ead an account o f it by
some phi l osopher who is not pr epar ed t o a dopt a di scr i mi nat or y at t i t ude
t oward pr oper t i es and who t hi nks t hat all t hi ngs ar e equal l y si mi l ar and
dissimilar t o one anot her .
Laws and Causation. The obs er vat i on t hat physi cs di scovers nat ur al
propert i es in t he cour se of di scoveri ng l aws ma y serve t o i nt r oduce our next
topic: t he anal ysi s o f what it is t o be a l aw of nat ur e. I agree wi t h Ar ms t r ong
that we need uni versal s, or at l east nat ur al pr oper t i es, in expl ai ni ng what
l awhood is, t hough I di sagree wi t h his account of how t hi s is so.
Ar ms t r ong' s t heor y, in its si mpl est f or m, 25 hol ds t hat what makes cer t ai n
regularities l awful are second- or der states of af f ai r s N( F, G) in whi ch t he t wo
ordinary, first-order universals F and G are related by a cert ai n dyadi c second-
order uni ver sal N. I t is a cont i ngent mat t er whi ch uni versal s are t hus r el at ed
by t he l awnmaker N. But it is necessary - - and necessar y simpliciter, not
just nomol ogi cal l y necessary - t hat i f N( F, G) obt ai ns, t hen F and G
are const ant l y conj oi ned. Ther e is a necessary connect i on bet ween t he
second-order st at e of af f ai r s N( F, G) and t he fi rst -order l awful r egul ar i t y
Vx(FxD Gx); and likewise bet ween t he conj unct i ve st at e of af f ai r s N( F, G)
& Fa and its necessary consequence Ga.
A paral l el t heor y coul d be set up wi t h nat ur al pr oper t i es in pl ace o f
Ar mst r ong' s first- and second- or der uni versal s. I t woul d have ma n y of t he
attractive feat ures t hat Ar ms t r ong cl ai ms on behal f of his t heor y, but at least
Universals, II, pp. 148-157. A more developed form of the theory appears in D. M.
Armstrong, What Is a Law of Nature? (Cambridge University Press, 1983). Similar theories
have been proposed in Fred I. Dretske, 'Laws of Nature', Philosophy of Science 44 (1977)
pp. 248-268, and in Michael Tooley, 'The Nature of Laws', Canadian Journal of Philosophy
4 (1977) pp. 667-698.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
366 Ne w Wor k f o r a Theory o f Universals
one mer i t wo u l d be l ost . Fo r Ar ms t r o n g , t he l awf ul neces s i t at i on o f Ga by
Fa is a pur e l y l ocal ma t t e r : it i nvol ves onl y a, t he uni ver sal s F a n d G t hat
ar e pr es ent i n a, a n d t he s e c o n d - o r d e r l a wma k i n g uni ver s al t h a t is present
i n t u r n i n ( or bet ween) t hes e t wo uni ver sal s. I f we r epl ace t he uni ver sal s by
pr oper t i es , h o we v e r na t ur a l , t ha t l ocal i t y is l ost . F o r pr oper t i es ar e classes
wi t h t hei r me mb e r s h i p s pr ead a r o u n d t he wor l ds , a nd ar e n o t whol l y present
i n a. But I d o n o t t h i n k t hi s a concl us i ve obj e c t i on, f o r o u r i nt ui t i ons of
l ocal i t y o f t e n s eem t o l ead us as t r ay. Th e sel ect i ve r egul ar i t y t h e o r y I shall
s hor t l y a dvoc a t e al so sacri fi ces l ocal i t y, as does a n y r egul ar i t y t he or y o f law.
Wh a t l eads me ( wi t h s ome r egr et ) t o r ej ect Ar ms t r o n g ' s t he or y, whet her
wi t h uni ver sal s o r wi t h na t ur a l pr oper t i es , is t ha t I f i nd its necessar y
c onne c t i ons uni nt el l i gi bl e. Wh a t e v e r N ma y be, I c a n n o t see h o w it coul d
be abs ol ut el y i mpos s i bl e t o ha ve N( F , G) a n d Fa wi t h o u t Ga. ( Unl ess N j ust
is cons t ant c onj unc t i on, or cons t ant c onj unc t i on pl us s omet hi ng else, i n which
case Ar ms t r o n g ' s t h e o r y t ur ns i nt o a f o r m o f t he r egul ar i t y t he or y he rejects. )
Th e my s t e r y is s o me wh a t hi dde n b y Ar ms t r o n g ' s t e r mi nol ogy. He uses
' neces s i t at es ' as a n a me f or t he l a wma k i n g uni ver s al N; a nd wh o wo u l d be
s u r p r i s e d ' t o hear t h a t i f F ' neces s i t at es ' G a n d a has F, t he n a mu s t have
G? But I s ay t ha t N deser ves t he n a me o f ' neces s i t at i on' o n l y i f, s ome how,
it r eal l y can ent er i nt o t he r equi si t e neces s ar y c onne c t i ons . I t c a n' t ent er i nt o
t h e m j us t by be a r i ng a na me , a n y mo r e t h a n one can have mi g h t y biceps
j us t b y bei ng cal l ed ' Ar ms t r o n g ' .
I a m t e mp t e d t o c o mp l a i n i n Hu me a n f a s hi on o f al l eged necessar y
c onne c t i ons be t we e n di st i nct exi st ences, especi al l y wh e n f i r s t - or der st at es of
af f ai r s i n t he pa s t s uppos e dl y j o i n wi t h s e c o n d - o r d e r st at es o f af f ai r s t o
necessi t at e f i r s t - or der st at es o f af f ai r s i n t he f ut ur e. Th a t c o mp l a i n t is not
cl ear l y r i ght : t he s har i ng o f Uni versal s det r act s f r o m t he di st i nct ness o f the
neces s i t at i ng a n d t he neces s i t at ed st at es o f af f ai r s . But I a m n o t appeased.
I c onc l ude t ha t neces s ar y c onne c t i ons can be uni nt el l i gi bl e even wh e n t hey
ar e s u p p o s e d t o o b t a i n be t we e n exi st ences t h a t ar e n o t cl ear l y a nd whol l y
di st i nct . 26
Th u s I d o not e ndor s e Ar ms t r o n g ' s wa y o f bui l di ng uni ver sal s, or
al t er nat i vel y nat ur al pr oper t i es, i nt o t he anal ysi s o f l a whood. I ns t ead I f avour
a r e gul a r i t y anal ysi s. But I need na t ur a l pr oper t i es even so.
Ce r t a i nl y n o t j us t a n y r e gul a r i t y is a l aw o f na t ur e . So me ar e acci dent al .
26
Armstrong's more developed theory in What Is a Law of Nature? complicates the picture
in two ways. First, the second order state of affairs N(F,G) is itself taken to be a universal,
and its presence in its instances detracts yet further from the distinctness of the necessitating
and the necessitated states of affairs. Second, all laws are defeasible, It is possible after all
to have N(F,G) and Fa without Ga, namely if we also have N(E&F,H) and Ea, where H
and G are incompatible. (The law that F' s are G's might be contingently indeafeasible, if
no such defeating state of affairs N(E&F,H) obtains; but no law has its indefeasibility built
in essentially.) It remains true that there are alleged necessary connections that I find
unintelligible, but they are more complicated than before. To necessitate a state of affairs,
we need not only the first- and second-order states of affairs originally considered, but also
a negative existential to the effect that there are no further states of affairs of the sort that
could act as defeaters.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 367
So an adequat e r egul ar i t y anal ysi s mus t be selective. Al so, an adequat e
analysis mus t be collective. I t mus t t r eat regul ari t i es not one at a t i me, but
rat her as candi dat es t o ent er i nt o i nt egr at ed syst ems. For a gi ven r egul ar i t y
might hol d ei t her as a l aw or acci dent al l y, dependi ng on whet her ot her
regularities obt ai n t hat can fit t oget her wi t h it in a sui t abl e syst em. (Thus
I rej ect t he i dea t hat l awhood consi st s of ' l awl i keness' pl us t r ut h. ) Fol l owi ng
Mill and Rams ey, 27 I t ake a sui t abl e syst em t o be one t hat has t he vi rt ues
we aspi re t o in our own t heor y- bui l di ng, and t hat has t hem t o t he great est
extent possi bl e gi ven t he way t he wor l d is. I t mus t be ent i rel y t rue; it mus t
be cl osed under st ri ct i mpl i cat i on; it mus t be as si mpl e in axi omat i s at i on as
it can be wi t hout sacrificing t oo muc h i nf or mat i on cont ent ; and it mus t have
as much i nf or mat i on cont ent as it can have wi t hout sacrificing t oo much
simplicity. A l aw is any r egul ar i t y t hat earns i ncl usi on in t he ideal syst em.
(Or, in case of ties, in every i deal syst em. ) The i deal syst em need not consi st
entirely of regul ari t i es; par t i cul ar fact s ma y gai n ent r y i f t hey cont r i but e
enough t o collective si mpl i ci t y and st rengt h. ( For i nst ance, cert ai n par t i cul ar
facts about t he Big Bang mi ght be st rong candi dat es. ) But onl y t he regularities
of t he syst em are t o count as laws.
We face an obvi ous pr obl em. Di f f er ent ways t o express t he s ame cont ent ,
using di f f er ent vocabul ar y, will di f f er in si mpl i ci t y. The pr obl e m can be put
in t wo ways, dependi ng on whet her we t ake our syst ems as consi st i ng of
proposi t i ons (classes of worl ds) or as consi st i ng of i nt er pr et ed sent ences. I n
the first case, t he pr obl e m is t hat a single syst em has di f f er ent degrees of
simplicity rel at i ve t o di f f er ent linguistic f or mul at i ons . I n t he second case,
the pr obl e m is t hat equi val ent syst ems, st ri ct l y i mpl yi ng t he ver y same
regularites, ma y di f f er in t hei r si mpl i ci t y. I n f act , t he cont ent of any syst em
whatever ma y be f or mul at ed ver y si mpl y i ndeed. Gi ven syst em S, let F be
a predi cat e t hat appl i es t o all and onl y t hi ngs at wor l ds wher e S hol ds. Take
Fas pri mi t i ve, and axi omat i se S (or an equi val ent t hereof) by t he single axi om
YxFx. I f ut t er si mpl i ci t y is so easily at t ai ned, t he i deal t heor y ma y as well
be as st r ong as possi bl e. Si mpl i ci t y and st r engt h needn' t be t r aded of f . Then
the ideal t heor y will i ncl ude (its si mpl e axi om will st ri ct l y i mpl y) all t r ut hs,
and af ort i ori all regul ari t i es. Then, af t er all, ever y r egul ar i t y will be a l aw.
That mus t be wr ong.
The r emedy, of course, is not t o t ol erat e such a perverse choi ce of pri mi t i ve
vocabulary. We shoul d ask how candi dat e systems compar e in simplicity when
each is f or mul at ed in t he si mpl est eligible way; or, i f we count di f f er ent
formul at i ons as di f f er ent syst ems, we shoul d dismiss t he ineligible ones f r om
candidacy. An appr opr i at e s t andar d of eligibility not f ar t o seek: let t he
John Stuart Mill, A System of Logic (Parker, 1843) Book III, Chapter IV, Section 1;
F. P. Ramsey, 'Universals of Law and of Fact', in his Foundations (Roufledge & Kegan
Paul, 1978). Ramsey regarded this theory of law as superseded by the different theory in
his 'General Propositions and Causality', also in Foundations, but I prefer his first thoughts
to his second. 1 present a theory of lawhood along the lines of Ramsey's earlier theory in
my Counterfactuals (Blackwell, 1973) pp. 73-75. A revision to that discussion is needed in
the probabilistic case, which I here ignore.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
368 New Work f or a Theory of Universals
primitive vocabulary that appears in the axioms refer only to perfectly natural
properties.
Of course, it remains an unsolved and difficult problem to say what
simplicity of a formulation is. But it is no longer the downright insoluble
problem that it would be if there were nothing to choose between alternative
primitive vocabularies.
(One might think also to replace strict implication by deducibility in some
specified calculus. But this second remedy seems unnecessary given the first,
and seems incapable of solving our problem by itself.)
If we adopt the remedy proposed, it will have the consequence that laws
will tend to be regularities involving natural properties. Fundamental laws,
those that the ideal system takes as axiomatic, must concern perfectly natural
properties. Derived laws that follow fairly straightforwardly also will tend
to concern fairly natural properties. Regularities concerning unnatural
properties may indeed be strictly implied, and should count as derived laws
if so. But they are apt to escape notice even if we someday possess a good
approximation to the ideal system. For they will be hard to express in a
language that has words mostly for not-too-unnatural properties, as any
language must. (See the next section.) And they will be hard to derive, indeed
they may not be finitely derivable at all, in our deductive calculi. Thus my
account explains, as Armstrong's does in its very different way, why the
scientific investigation of laws and of natural properties is a package deal;
why physicists posit natural properties such as the quark colours in order
to posit the laws in which those properties figure, so that laws and natural
properties get discovered together.
If the analysis of lawhood requires natural properties, then so does the
analysis of causation. It is fairly uncontroversial that causation involves laws.
That is so according to both of the leading theories of causation: the
deductive-nomological anaiysis, on which the laws are applied to the actual
course of events with the cause and effect present; and the counterfactual
analysis that I favour, on which the laws are applied to counterfactual
situations with the cause hypothesised away. These counterfactual alternatives
may need to break actual laws at the point where they diverge from actuality,
but the analysis requires that they evolve thereafter in accordance with the
actual laws. 2s
According to my counterfactual analysis, causation involves natural
properties in a second way too. We need the kind of counterfactuals that
avoid backtracking; else the analysis faces fatal counterexamples involving
epiphenomenai side-effects or cases of causal preemption. As I have already
noted, these counterfactuals are to be characterised in terms of divergent
28 See my 'Causation', Journal of Philosophy 70 (1973) pp. 556-567; reprinted in Ernest Sosa,
ed., Causation and Conditionals (Oxford University Press, 1975).
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 369
worlds, hence in t er ms of dupl i cat e initial worl d-segment s, hence in t er ms
of shared nat ur al propert i es.
Causat i on involves nat ur al pr oper t i es in yet anot her way. (Small wonder
that I came t o appr eci at e nat ur al propert i es af t er wor ki ng on t he analysis
of causat i on!) Causat i on hol ds bet ween events. Unless we distinguish genuine
from spuri ous events, we will be l eft with t oo many put at i ve causes. You
put a l ump of but t er on a skillet, and t he but t er melts. What event causes
this? Ther e is one event t hat we can call a movi ng of mol ecul es. It occurs
in the regi on where t he skillet is, j ust bef or e t he but t er melts. Thi s is an event
such t hat , necessarily, it occurs in a spat i ot empor al regi on onl y i f t hat region
contains rapi dl y movi ng molecules. Surel y this event is a cause of t he melting
of t he but t er .
Heat is t hat phenomenon, what ever it ma y be, t hat mani fest s itself in
certain fami l i ar charact eri st i c ways. Let us say: heat is t hat whi ch occupi es
the heat-role. (It won' t mat t er whet her we t ake t he defi ni t e descri pt i on plain,
as I pr ef er , or rigidified. ) In fact , but cont i ngent l y, it is mol ecul ar mot i on
that occupi es t he heat -rol e. It mi ght have been mol ecul ar nonmot i on, or
caloric fluid, or what you will. Now consi der an alleged second event , one
that we may call a havi ng- t he- occupant - of - t he- heat - r ol e. Thi s second event
occurs j ust when and where t he first does, in t he r egi on where t he hot skillet
is. It occurs t her e in vi rt ue of t he t wo facts (1) t hat t he skillet' s mol ecul es
are movi ng r api dl y, and (2) t hat t he regi on in quest i on is par t of a wor l d
where mol ecul ar mot i on is what occupies t he heat -rol e. But this second event
differs f r om the first. The necessary conditions f or its occurrence are di fferent .
Necessarily, it occurs in a regi on onl y i f t hat regi on cont ai ns what ever
phenomenon occupies t he heat -rol e in t he worl d of whi ch t hat regi on is part .
So in t hose worl ds where cal ori c fluid occupi es t he heat - r ol e and mol ecul ar
motion does not , t he first event occurs onl y in regi ons wi t h mol ecul ar mot i on
whereas t he second occurs onl y in regions with cal ori c fluid.
Cert ai nl y t he first event causes t he mel t i ng of t he but t er , but shall we say
that t he second event does so as well? No; t hat seems t o mul t i pl y causes
beyond bel i ef by pl ayi ng a verbal t ri ck. But i f t her e real l y are t wo events
here, I cannot see why t he second has less of a cl ai m t han t he first t o be a
cause of t he mel t i ng of t he but t er . It is out of t he quest i on t o say t hat t he
first and t he second events are one and t he same - t hen this one event woul d
have di f f er ent condi t i ons of occur r ence f r om itself. The best sol ut i on is t o
deny t hat t he alleged second event is a genuine event at all. I f it isn' t, of course
it can' t do any causing.
Why is t he first event genui ne and t he second spur i ous? Compar e t he
properties i nvol ved: cont ai ni ng r api dl y movi ng mol ecul es versus cont ai ni ng
whatever phe nome non occupi es t he heat -rol e. (I mean t hese as propert i es
of t he spat i ot empor al regi on; ot her t r eat ment s of events woul d t ake i nst ead
the cor r espondi ng pr oper t i es of t he skillet, but my poi nt woul d still appl y. )
The first is a fairly nat ural , intrinsic pr oper t y. The second is highly disjunctive
and extrinsic. For all sorts of di f f er ent phenomena coul d occupy t he heat -
role; and whet her t he phe nome non goi ng on in a r egi on occupi es t he rol e
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
370 New Work for a Theory of Universab
d e p e n d s n o t o n l y o n wh a t goes o n i n t h e r e g i o n b u t a l s o o n wh a t goes on
e l s e whe r e i n t he s a me wo r l d . Th u s t h e d i s t i n c t i o n b e t we e n mo r e a n d less
n a t u r a l p r o p e r t i e s gi ves me t h e d i s t i n c t i o n b e t we e n g e n u i n e a n d s pur i ous
e ve nt s t h a t I ne e d i n o r d e r t o d i s o wn a n o v e r a b u n d a n c e o f c a us e s . I f a
p r o p e r t y i s t o o u n n a t u r a l , i t i s i ne f f i c a c i ous i n t he s ens e t h a t i t c a n n o t figure
i n t he c o n d i t i o n s o f o c c u r r e n c e o f t h e e ve nt s t h a t c a us e t hi ngs . 29
The Cont ent of Language and Thought. Hi l a r y P u t n a m ha s gi ven an
a r g u me n t whi c h he r e g a r d s as a r e f u t a t i o n o f a ' r a d i c a l l y n o n - e p i s t e mi c ' view
o f t r u t h , b u t whi c h I r e g a r d r a t h e r as a reductio a g a i n s t P u t n a m' s pr emi s es . 3
I n p a r t i c u l a r , i t r e f ut e s hi s a s s u mp t i o n t h a t ' we i n t e r p r e t o u r l a n g u a g e s or
n o t h i n g d o e s ' ( ' Mo d e l s a n d Re a l i t y ' , p. 482) so t h a t a n y c o n s t r a i n t on
r e f e r e nc e mu s t be e s t a b l i s h e d b y o u r o wn s t i p u l a t i o n i n l a n g u a g e o r t hought .
Ga r y Me r r i l l ha s s u g g e s t e d t h a t P u t n a m ma y b e a n s we r e d b y a p p e a l t o a
c o n s t r a i n t t h a t d e p e n d s o n a n o b j e c t i v e s t r u c t u r e o f p r o p e r t i e s a n d r el at i ons
i n t h e wo r l d . 31 I a gr e e , a n d f i nd he r e a n o t h e r p o i n t a t whi c h we ne e d nat ur al
p r o p e r t i e s .
P u t n a m' s a r g u me n t , as I u n d e r s t a n d i t , is as f o l l o ws . Fi r s t , s u p p o s e t hat
t he o n l y c o n t r a i n t o n i n t e r p r e t a t i o n o f o u r l a n g u a g e ( or p e r h a p s o u r l anguage
o f t h o u g h t ) i s gi ve n b y a d e s c r i p t i o n t h e o r y o f r e f e r e n c e o f a g l o b a l and
f ut ur i s t i c s or t . An ' i n t e n d e d i n t e r p r e t a t i o n ' i s a n y i n t e r p r e t a t i o n t h a t satisfies
a c e r t a i n b o d y o f t h e o r y : viz. t h e i d e a l i s e d d e s c e n d a n t o f o u r c u r r e n t t ot al
t h e o r y t h a t wo u l d e me r g e at t h e e n d o f i n q u i r y , a n i de a l t h e o r y r ef i ned to
p e r f e c t i o n u n d e r t he g u i d a n c e o f al l n e e d e d o b s e r v a t i o n a n d o u r best
t h e o r e t i c a l r e a s o n i n g . I f s o, i n t e n d e d i n t e r p r e t a t i o n s a r e s ur pr i s i ngl y
a b u n d a n t . F o r any wo r l d c a n s a t i s f y any t h e o r y ( i de a l o r not ) , a n d c a n do
so i n c o u n t l e s s v e r y d i f f e r e n t wa ys , p r o v i d e d o n l y t h a t t he wo r l d is n o t t oo
s ma l l a n d t h e t h e o r y i s c o n s i s t e n t . Be y o n d t h a t , i t d o e s n ' t ma t t e r wh a t the
wo r l d is l i ke o r wh a t t h e t h e o r y s ays . He n c e we ha ve r a d i c a l i n d e t e r mi n a c y
o f r e f e r e n c e . An d we h a v e t he c o i n c i d e n c e t h a t P u t n a m we l c o me s bet ween
s a t i s f a c t i o n u n d e r al l i n t e n d e d i n t r e p r e t a t i o n s a n d ' e p i s t e mi c t r u t h ' . F o r the
i de a l t h e o r y is t h e wh o l e o f ' e p i s t e mi c t r u t h ' , t h e i n t e n d e d i n t e r p r e t a t i o n s are
j u s t t h o s e i n t e r p r e t a t i o n s o f o u r l a n g u a g e t h a t s a t i s f y t h e i de a l t h e o r y , and
z9 See the discussion of impotence of dispositions in Elizabeth W. Prior, Robert Pargetter,
and Frank Jackson, 'Three Theses About Dispositions', American Philosophical Quarterly
19 (1982) pp. 251-257. If a disposition is not identified with its actual basis, there is a threat
of multiplication of putative causes similar to that in my example. We would not wish to
say that the breaking of a struck glass is caused both by its fragility and by the frozen-in
stresses that are the basis thereof; and if forced to choose, we should choose the latter. I
suggest that the fragility is inefficacious because it is too unnatural a property, too disjunctive
and extrinsic, to figure in the conditions of occurence of any event.
3o Hilary Putnam, 'Realism and Reason', in his Meaning and the Moral Sciences (Routledge
& Kegan Paul, 1978), and 'Models and Reality', Journal of Symbolic Logic 45 (1980)
pp. 464-482. The reader is warned that the argument as I present it may not be quite as
Putnam intended it to be. For I have made free in reading between the lines and in restating
the argument in my own way.
31 G. H. Merrill, 'The Model-Theoretic Argument Against Realism', Philosophy of Science
47 (1980) pp. 69-81.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 371
(unless t he wor l d is t oo smal l or i deal t heor y is i nconsi st ent ) t her e are s ome
such i nt er pr et at i ons.
I t ake t hi s t o r ef ut e t he supposi t i on t hat t her e are no f ur t her const r ai nt s
on reference. But Pu t n a m asks: how coul d t her e be a f ur t her const r ai nt ?
How coul d we ever est abl i sh it? By st i pul at i on, by sayi ng or t hi nki ng
something. But what ever we say or t hi nk will be in l anguage (or l anguage
of t hought ) t hat suf f er s f r om r adi cal i ndet er mi nacy of i nt er pr et at i on. For
the savi ng const r ai nt will not be t her e unt i l we succeed in est abl i shi ng it. So
the at t empt ed st i pul at i on mus t fail. The mos t we can do is t o cont r i but e a
new chapt er t o cur r ent and i deal t heor y, a chapt er consi st i ng o f what ever
we said or t hought in our st i pul at i on. And t hi s new t heor y goes t he way of
all t heor y. So we cannot est abl i sh a f ur t her const r ai nt ; and ' we i nt er pr et our
language or not hi ng does' ; so t her e cannot be any f ur t her const r ai nt . We
cannot l i ft oursel ves by our boot s t r aps , so we mus t still be on t he gr ound.
Indeed we cannot lift oursel ves by our boot s t r aps , but we are of f t he
ground, so t her e mus t be anot her way t o fly. Our l anguage does have a fai rl y
determinate i nt erpret at i on (a Moor ean fact !) so t here mus t be some const rai nt
not cr eat ed ex nihilo by our st i pul at i on.
What can it be? Ma ny phi l osopher s woul d suggest t hat it is s ome sort of
causal const r ai nt . I f so my case is made, gi ven my ar gument s in t he pr evi ous
section: we need nat ur al pr oper t i es t o expl ai n det er mi nacy of i nt er pr et at i on.
But I doubt t hat it real l y is a causal const r ai nt , f or I am i ncl i ned t o t hi nk
that t he causal aspect of r ef er ence is est abl i shed by what we say and t hi nk.
Thus: I t hi nk o f a t hi ng as t hat whi ch I a m causal l y acquai nt ed wi t h in such-
and-such way, per haps per cept ual l y or per haps t hr ough a channel of
acquaintance t hat i nvol ves t he nami ng o f t he t hi ng and my pi cki ng up of
the name. I r ef er t o t hat t hi ng in my t hought , and der i vat i vel y in l anguage,
because it is t he t hi ng t hat fits t hi s causal and egocent ri c descri pt i on ext r act ed
from my t heor y o f t he wor l d and of my pl ace in t he wor l d. 32
I woul d i nst ead pr opos e t hat t he savi ng const r ai nt concer ns t he r ef er ent
- not t he r ef er r er , and not t he causal channel s bet ween t he t wo. I t t akes
two to ma ke a r ef er ence, and we will not fi nd t he const r ai nt i f we l ook f or
it always on t he wr ong side o f t he r el at i onshi p. Ref er ence consi st s in par t
of what we do in l anguage or t hought when we refer, but in par t it consi st s
in eligibility o f t he r ef er ent . And t hi s eligibility t o be r ef er r ed t o is a mat t er
of nat ural pr oper t i es.
That is t he suggest i on Merri l l of f er s. ( He of f er s it not as his own vi ew,
but as what opponent s of Pu t n a m ought t o say; and I gr at ef ul l y accept t he
offer.) I n t he si mpl est case, suppose t hat t he i nt er pr et at i on of t he l ogi cal
vocabul ary s ome how t akes car e of itself, t o reveal a s t andar d fi rst -order
language whose nonl ogi cal voc a bul a r y consi st s ent i rel y o f pr edi cat es. The
parts o f t he wor l d compr i s e a domai n; and sets, sets o f pai r s . . . . . f r o m t hi s
domain are pot ent i al ext ensi ons f or t he pr edi cat es. Now suppose we have
32 See Stephen Schiffer, 'The Basis of Reference', Erkenntnis 13 (1978) pp. 171-206.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
372
New Work f or a Theory of Universals
an al l - or - not hi ng division of pr oper t i es i nt o nat ur al and unnat ur al . Say that
a set f r om t he domai n is eligible t o be t he ext ensi on of a one-pl ace predicate
i f f its member s are j ust t hose things in t he domai n t hat share some natural
pr oper t y; and likewise f or many- pl ace predi cat es and nat ur al rel at i ons. An
eligible interpretation is one t hat assigns none but eligible ext ensi ons t o the
predi cat es. A so-called ' i nt ended' i nt er pr et at i on is an eligible i nt erpret at i on
t hat satisfies t he i deal t heor y. (But t he name is misleading: it is not t o be
said t hat our intentions establish t he const rai nt requiring eligibility. That way
lies t he futile boot st r ap- t uggi ng t hat we must avoi d. ) Then i f t he natural
pr oper t i es are sparse, t her e is no r eason t o expect any over abundance of
i nt ended i nt er pr et at i ons. Ther e ma y even be none. Even ideal t heor y runs
t he risk of beings unsatisfiable, save in ' uni nt ended' ways. Because satisfaction
is not guar ant eed, we accompl i sh somet hi ng i f we manage t o achi eve it by
maki ng a good fit bet ween t heor y and t he wor l d. All this is as it shoul d be.
The pr oposal calls f or refinement. First, we need t o provi de f or richer forms
of l anguage. In this we can be gui ded by fami l i ar t ransl at i ons, f or instance
bet ween modal l anguage wi t h hi gher - or der quant i f i cat i on and first-order
l anguage t hat explicitly ment i ons possibilia and classes bui l t up f r om them.
Second, it will not do t o t ake naturalness of properties as all-or-nothing. Here,
above all, we need t o make nat ural ness - and hence eligibility - a
compar at i ve mat t er , or a mat t er of degree. Ther e are salient shar p lines, but
not in t he ri ght places. Ther e is t he line bet ween t he per f ect l y natural
pr oper t i es and all t he rest, but surel y we have predi cat es f or much-less-than-
perfect l y nat ural propert i es. Ther e is t he line bet ween propert i es t hat are and
t hat are not finitely anal ysabl e in t erms of per f ect l y nat ur al propert i es, but
t hat lets in enough highly unnat ur al pr oper t i es t hat it t hr eat ens not t o solve
our pr obl em. We need gradat i ons; and we need some give and t ake between
t he eligibility of r ef er ent s and t he ot her fact ors t hat make f or ' intendedness' ,
not abl y sat i sfact i on of appr opr i at e bits of t heor y. (Ideal t heor y, i f we keep
as much of Put nam' s st ory as we can. ) Grueness is not an absolutely ineligible
r ef er ent (as witness my r ef er ence t o it j ust now) but an i nt er pr et at i on that
assigns it is t o t hat extent i nferi or t o one t hat assigns blueness instead. Ceteris
paribus, the latter is the ' intended' one, just because it does bet t er on eligibility.
Nat ural ness of propert i es makes f or di fferences of eligibility not only
among t he propert i es t hemsel ves, but also among t hi ngs. Compar e Bruce
wi t h t he cat -shaped chunk of mi scel l aneous and ever-changi ng mat t er that
follows him ar ound, always a few steps behi nd. The f or mer is a highly eligible
r ef er ent , t he l at t er is not . (I haven' t succeeded in r ef er r i ng t o it, f or I didn't
say j ust whi ch such chunk ' it' was t o be. ) That is because Bruce, unlike the
cat -shaped chunk, has a boundar y well demar cat ed by di fferences in highly
nat ural propert i es. Wher e Bruce ends, t here t he density of mat t er, t he relative
abundance of t he chemi cal el ement s . . . . abr upt l y change. Not so f or the
chunk. Bruce is also much mor e of a locus of causal chains t han is t he chunk;
this t oo traces back t o nat ural properties, by t he considerations of t he previous
sect i on. Thus nat ural ness of pr oper t i es sets up di st i nct i ons among things.
The reverse happens also. Once we are away f r om t he per f ect l y natural
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 373
properties, one t hi ng t hat makes f or nat ural ness of a pr oper t y is t hat it is
a pr oper t y bel ongi ng exclusively t o wel l -demarcat ed t hi ngs.
You mi ght well pr ot est t hat Put nam' s pr obl em is mi sconcei ved, wher ef or e
no need has been demonst r at ed f or resources t o solve it. Put na m seems t o
conceive of l anguage ent i rel y as a r eposi t or y of t heor y, and not at all as a
practice of social i nt eract i on. We have t he l anguage of t he encycl opedi a, but
where is t he l anguage of t he pub? Wher e are t he communi cat i ve i nt ent i ons
and t he mut ual expect at i ons t hat seem t o have so much t o do with what we
mean? I n fact , where is t hought ? It seems t o ent er t he pi ct ure, i f at all, onl y
as t he special case where t he l anguage t o be i nt er pr et ed is har d- wi r ed,
unspoken, hi dden, and all t oo conj ect ur al .
I t hi nk t he poi nt is well t aken, but I t hi nk it doesn' t mat t er . I f t he pr obl em
of i nent i onal i t y is ri ght l y posed t her e will still be a t hr eat of radi cal
i ndet ermi nacy, t her e will still be a need f or saving const rai nt s, t her e will still
be a r emedy anal ogous t o Merrill' s suggested answer t o Put nam, and t her e
will still be a need f or nat ur al propert i es.
Set l anguage aside and consi der i nst ead t he i nt er pr et at i on of t hought .
(Aft erward we can hope t o i nt er pr et t he subj ect ' s l anguage in t er ms of his
beliefs and desires regardi ng verbal communi cat i on with ot hers. ) The subj ect
is in vari ous states, and coul d be in vari ous ot hers, t hat are causally rel at ed
to each ot her, t o t he subject' s behavi our , and t o t he near by envi r onment t hat
stimulates his senses. These states fit i nt o a f unct i onal organi sat i on, t hey
occupy cer t ai n causal roles. (Most likely t hey are states of t he brai n. Maybe
they i nvol ve somet hi ng t hat is l anguage-l i ke but hard-wi red, maybe not . But
the nat ur e of t he states is besi de t he poi nt . ) The states have t hei r f unct i onal
roles in t he subj ect as he now is, and in t he subj ect as he is at ot her t i mes
and as he mi ght have been under ot her ci rcumst ances, and even in ot her
creatures of t he same ki nd as t he subj ect . Gi ven t he f unct i onal roles of t he
states, t he pr obl em is t o assign t hem cont ent . Pr oposi t i onal cont ent , some
would say; but I woul d agree onl y i f t he pr oposi t i ons can be t aken as
egocentric ones, and I t hi nk an ' egocent ri c pr oposi t i on' is simply a pr oper t y.
States i ndexed by cont ent can be identified as a bel i ef t hat this, a desire f or
that, a per cept ual experi ence of seeming t o conf r ont so-and-so, an i nt ent i on
to do such-and-such. (But not all or di nar y ascri pt i ons of at t i t udes merel y
specify t he cont ent of t he subj ect ' s states. Fr ed and Ted mi ght be alike in
the funct i onal roles of t hei r states, and hence have states with the same cont ent
in t he nar r owl y psychol ogi cal sense t hat is my present concern, and hence
believe alike e.g. by each believing hi msel f t o have hear d of a pr et t y t own
named ' Cast l emai ne ~. Yet t hey mi ght be acquai nt ed vi a t hat name wi t h
different t owns, at opposi t e ends of t he ear t h, so t hat Fr ed and not Ted
believes t hat Cast l emai ne, Vi ct ori a, is pr et t y. ) The pr obl em of assigning
content t o f unct i onal l y charact eri sed states is t o be solved by means of
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
374 New Work f or a Theory of Universals
const rai ni ng principles. For emost among these are principles of fit. I f a state
is t o be i nt er pr et ed as an i nt ent i on t o raise one' s hand, it had bet t er typically
cause t he hand t o go up. I f a st at e (or compl ex of states) is t o be interpreted
as as a syst em of beliefs and desires - or bet t er , degrees of bel i ef and desire
- accor di ng t o whi ch raising one' s hand woul d be a good means t o one's
ends, and i f anot her st at e is t o be i nt er pr et ed as an i nt ent i on t o raise one's
hand, t hen t he f or mer had bet t er t ypi cal l y cause t he l at t er. Likewise on the
i nput side. A st at e t ypi cal l y caused by r ound things bef or e t he eyes is a good
candi dat e f or i nt erpret at i on as t he visual experience of conf r ont i ng something
r ound; and its t ypi cal i mpact on t he states i nt er pr et ed as systems of belief
ought t o be i nt er pr et ed as t he exogenous addi t i on of a bel i ef t hat one is
conf r ont i ng somet hi ng r ound, wi t h what ever adj ust ment t hat addi t i on calls
f or .
So f ar , so good. But it seems clear t hat pr epost er ous and perverse
mi si nt er pr et at i on s coul d nevertheless coher e, coul d manage t o fit the
f unct i onal roles of t he states because mi sassi gnment of cont ent at one point
compensat es f or mi sassi gnment at anot her . Let us see j ust how this could
happen, at least under an oversi mpl i fi ed pi ct ur e of i nt er pr et at i on as follows.
An i nt er pr et at i on is given by a pai r of f unct i ons C and V. C is a probability
di st ri but i on over t he worl ds, r egar ded as encapsul at i ng t he subject's
dispositions t o f or m beliefs under t he i mpact of sensory evidence: i f a stream
of evi dence specified by pr oposi t i on E woul d put t he subj ect i nt o a t ot al state
S - f or short , i f E yi el ds S - we i nt er pr et S t o consi st in par t of t he belief
system given by t he pr obabi l i t y di st ri but i on C( - / E) t hat comes f r om C by
condi t i onal i si ng on E. V is a f unct i on f r om worl ds t o numeri cal desirability
scores, r egar ded as encapsul at i ng t he subj ect ' s basic values: i f E yields S,
we i nt er pr et S t o consist in par t of t he syst em of desires given by t he C(-/ E)-
expect at i ons of V. Say t hat C and V rationalise behavi our B af t er evidence
E i f f t he system of desires given by t he C( - / E) - expect at i ons of V ranks B
at least as high as any al t ernat i ve behavi our . Say t hat C and Vf i t iff, for
any evi dence-speci fyi ng E, E yields a st at e t hat woul d cause behaviour
rat i onal i sed by C and V af t er E. That is our onl y const rai ni ng prilxciple of
fit. ( Wher e di d t he ot hers go? -- We bui l t t hem i nt o t he definitions whereby
C and V encapsul at e an assi gnment of cont ent t o vari ous states. ) Then any
t wo i nt erpret at i ons t hat always rat i onal i se t he same behavi our af t er t he same
evi dence must fit equal l y well. Call t wo worl ds equi val ent i f f t hey are alike
in respect of t he subj ect ' s evi dence and behavi our , and not e t hat any decent
wor l d is equi val ent i nt er alia t o hor r endous l y count er i nduct i ve worl ds and
t o worl ds where everyt hi ng unobser ved by t he subj ect is hor r endousl y nasty.
Fi t depends on t he t ot al of C f or each equi val ence class, and on t he C-
expect at i on of V wi t hi n each class, but t hat is all. Wi t hi n a class, it makes
no di f f er ence whi ch wor l d gets which pai r of values of C and V. We can
i nt er change equi val ent worl ds ad lib and preserve fit. So, given any fitting
and r easonabl e i nt er pr et at i on, we can t r ans f or m it i nt o an equal l y fitting
perverse i nt erpret at i on by swapping equivalent worlds ar ound so as t o enhance
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Le wi s 375
the pr obabi l i t i es o f count er i nduct i ve worl ds, or t he desi rabi l i t i es of nast y
worlds, or bot h. Qu o d e r at d e mo n s t r a n d u m.
(My simplifications were dire: I left out t he egocent ri ci t y of bel i ef and desire
and evi dence, t he causal aspect of r at i onal i sed behavi our , t he rol e of
intentions, change of basi c val ues, l i mi t at i ons of l ogi cal compet ence . . . . .
But I doubt t hat t hese omi ssi ons mat t er t o my concl usi on. I conj ect ur e t hat
if t hey were r emedi ed, we coul d still t r a ns f or m r easonabl e i nt er pr et at i ons
into per ver se ones in a way t hat preserves fit.)
I f we rel y on pri nci pl es o f fit t o do t he whol e j ob, we can expect radi cal
indeterminacy of i nt erpret at i on. We need furt her const rai nt s, of t he sort called
principles of (sophi st i cat ed) char i t y, or of ' humani t y' . 33 Such pri nci pl es call
for i nt erpret at i ons accor di ng t o whi ch t he subj ect has at t i t udes t hat we woul d
deem r easonabl e f or one who has lived t he life t hat he has l i ved. (Unl i ke
principles of cr ude char i t y, t hey call f or i mput at i ons of er r or i f he has lived
under decept i ve condi t i ons. ) These pri nci pl es select a mong confl i ct i ng
interpretations t hat equal l y well c onf or m t o t he pri nci pl es of fit. They i mpose
a pr i or i - - al bei t def easi bl e - pr es umpt i ons about what sort s of t hi ngs are
apt t o be bel i eved and desi red; or r at her , about what di sposi t i ons t o devel op
beliefs and desires, what i nduct i ve bi ases and basi c val ues, s omeone ma y
rightly be i nt er pr et ed t o have.
It is her e t hat we need nat ur al pr oper t i es. The pri nci pl es of char i t y will
impute a bi as t owar d bel i evi ng t hat t hi ngs are green r at her t han grue, t owar d
having a basi c desi re f or l ong life r at her t han f or l ong-l i fe-unl ess-one-was
bor n- on- Monday- and- i n- t hat - cas e- l i f e- f or - an- even- number - of - weeks . I n
short, t hey will i mput e eligible cont ent , where ineligibility consi st s in severe
unnat ural ness of t he pr oper t i es t he subj ect supposedl y believes or desires or
intends hi msel f t o have. They will i mput e ot her t hi ngs as well, but it is t he
imputed eligibility t hat mat t er s t o us at pr esent .
Thus t he t hr eat of radi cal i ndet er mi nacy in t he assi gnment of cont ent t o
thought is f ended of f . The savi ng const r ai nt concer ns t he cont ent - not t he
thinker, and not any channel s bet ween t he t wo. I t t akes t wo t o i ndex st at es
with cont ent , and we will not find t he const r ai nt i f we l ook f or it al ways on
the wr ong side of t he r el at i onshi p. Believing t hi s or desi ri ng t hat consi st s
in par t in t he f unct i onal rol es o f t he st at es wher eby we bel i eve or desire, but
in par t it consi st s in t he eligibility of t he cont ent . And t hi s eligibility t o be
thought is a mat t er , in par t , of nat ur al pr oper t i es.
Consi der t he puzzl e wher eby Kr i pke illustrates Wi t t genst ei n' s par adox t hat
'no course of act i on coul d be det er mi ned by a rul e, because every cour se of
action can be ma de out t o accor d wi t h t he rul e' . 34 A wel l -educat ed per s on
working ar i t hmet i c pr obl ems i nt ends t o p e r f o r m addi t i on when he sees t he
' +' sign. He does not i nt end t o pe r f or m quaddi t i on, which is j ust like addi t i on
33 See my 'Radical Interpretation', Synthese 23 (1974) pp. 331-344; and Richard E. Grandy,
'Reference, Meaning and Belief', Journal of Philosophy 70 (1973) pp. 439-452.
34 See Saul A. Kripke, 'Wittgenstein on Rules and Private Language: An Elementary Exposition',
in Irving Block, ed., Perspectives on Wittgenstein (Blackwell, 1981).
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
376 New Work f or a Theory of Universals
f or small number s but whi ch yields t he answer 5 i f any o f t he number s to
be quadded exceeds a cert ai n bound. Wher ef or e does he i nt end t o add and
not t o quadd? What ever he says and what ever is wri t t en in his br ai n can be
perversel y (mi s)i nt erpret ed as i nst ruct i ng hi m t o quadd. And it is not enough
t o say t hat his br ai n st at e is t he causal basis of a di sposi t i on t o add. Perhaps
it isn' t. Per haps i f a test case arose he woul d abandon his i nt ent i on, he would
nei t her add nor quadd but i nst ead woul d put his home wor k aside and
compl ai n t hat t he pr obl ems are t oo har d.
The nai ve sol ut i on is t hat addi ng means going on in t he same way as before
when t he number s get big, whereas quaddi ng means doi ng something
di f f er ent ; t here is not hi ng present in t he subj ect t hat const i t ut es an intention
t o do di f f er ent things in di f f er ent cases; t her ef or e he i nt ends addi t i on, not
quaddi t i on. We shoul d not s cof f at this nai ve response. It is t he correct
sol ut i on t o t he puzzle. But we must pay t o regai n our naivet6. Our theory
of propert i es must have adequat e resources t o somehow r at i f y t he judgement
t hat i nst ances of addi ng are all alike in a way t hat instances of quadding
ar e not . The pr oper t y of addi ng is not per f ect l y nat ur al , of cour se, not on
a par wi t h uni t charge or spheri cal i t y. And t he pr oper t y of quaddi ng is not
per f ect l y unnat ur al . But quaddi ng is worse by a di sj unct i on. So quaddition
is t o t hat ext ent less of a way t o go on doi ng t he same, and t her ef or e it is
t o t hat ext ent less of an eligible t hi ng t o i nt end t o do.
It' s not t hat you coul dn' t possi bl y i nt end t o quadd. You coul d. Suppose
t hat t oda y t her e is as much basis as t her e ever is t o i nt er pr et you as intending
t o add and as meani ng addi t i on by your wor d ' addi t i on' and quaddi t i on by
' quaddi t i on' ; and t omor r ow you say t o your sel f in so many wor ds that it
woul d be f un t o t ease t he phi l osopher s by t aki ng up quaddi t on hencefort h,
and you make up your mi nd t o do it. But you have t o go out of your way.
Addi ng and quaddi ng ar en' t on a par . To i nt end t o add, you need onl y have
states t hat woul d fit ei t her i nt er pr et at i on and leave it t o char i t y t o decree
t hat you have t he mor e eligible i nt ent i on. To i nt end t o quadd, you must say
or t hi nk somet hi ng t hat creates difficulties of fit f or t he mor e eligible intention
and t her eby defeat s t he pr es umpt i on in its f avour . You must do something
t hat , t aki ng principles of fit and pr esumpt i ons of eligibility and other
pri nci pl es of char i t y t oget her , tilts t he bal ance in f avour of an interpretation
on whi ch you i nt end t o quadd. How i roni c t hat we were wor r i ed t o find
not hi ng posi t i ve t o settle t he mat t er in f avour of addi t i on! For t he lack of
anyt hi ng positive t hat points either way j ust is what it takes t o f avour addition.
Quaddi t i on, bei ng less nat ur al and eligible, needs somet hi ng positive in its
f avour . Addi t i on can wi n by def aul t .
What is t he status of t he pri nci pl es t hat const r ai n i nt er pr et at i on, in
part i cul ar t he chari t abl e pr esumpt i on in f avour of eligible cont ent ? We must
shun several mi sunder st andi ngs. It is not t o be said (1) t hat as a contingent
psychol ogi cal fact , t he cont ent s of our states t ur n out t o be fai rl y eligible,
we most l y believe and desire oursel ves t o have not - t oo- unnat ur al properties.
Still less shoul d it be said (2) t hat we shoul d dari ngl y pr esuppose this in our
i nt er pr et i ng of one anot her , even i f we haven' t a shred of evi dence for it.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9
David Lewis 377
Nor should it be said (3) that as a contingent psychological fact we turn out
to have states whose content involves some properties rather t han others,
and that is what makes it so that the former properties are more natural.
(This would be a psychologistic theory of naturalness.) The error is the same
in all three cases. It is supposed, wrongly as I think, that the problem of
interpretation can be solved without bringing to it the distinction between
natural and unnatural properties; so that the natural properties might or might
not turn out to be the ones featured in the content of t hought according to
the correct solution, or so that they can afterward be defined as the ones
that are so featured. I think this is overoptimistic. We have no notion how
to solve the problem of interpretation while regarding all properties as equally
eligible to feature in content. For that would be to solve it without enough
contraints. Only if we have an independent, objective distinction among
properties, and we impose the presumption in favour of eligible content a
priori as a constitutive constraint, does the problem of interpretation have
any solution at all. If so, then any correct solution must automatically respect
the presumption. There's no contingent fact of psychology here to be believed,
either on evidence or daringly.
Compare our selective and collective t heory of l awhood: l awhood of a
regularity just consists in its fitting into an ideally high-scoring system, so
it's inevitable that laws turn out to have what it takes to make for high scores.
Likewise, I have suggested, cont ent hood just consists in getting assigned by
a high-scoring interpretation, so it's inevitable that contents tend to have what
it takes to make for high scores. And in bot h cases, I' ve suggested that part
of what it takes is naturalness of the properties involved. The reason natural
properties feature in the contents of our attitudes is that naturalness is part
of what it is to feature therein. It's not that we're built to take a special interest
in natural properties, or that we confer naturalness on properties when we
happen to take an interest in them. 35
Princeton University Received August 1982
Revised May 1983
3~ I am indebted to comments by Gilbert Harman, Lloyd Humberstone, Frank Jackson, Mark
Johnston, Donald Morrison, Kim Sterelny, and others; and especially to discussion and
correspondence with D. M. Armstrong over several years, without which I might well have
believed to this day that set theory applied to possibilia is all the theory of properties that
anyone could ever need.
D
o
w
n
l
o
a
d
e
d

B
y
:

[
C
o
l
u
m
b
i
a

U
n
i
v
e
r
s
i
t
y
]

A
t
:

0
9
:
5
1

2
9

S
e
p
t
e
m
b
e
r

2
0
0
9

You might also like