Vous êtes sur la page 1sur 52

Appeared in International Studies in the Philosophy of Science, 17, 3, October 2003, pp. 225-244.

Scientific representation: Against similarity and isomorphism


MAURICIO SURE!
"epartment of #ogic and Philosophy of Science!
Uni$ersidad Complutense de Madrid! Spain%
A&stract:
I argue against theories that attempt to reduce scientific representation to similarity or
isomorphism% 'hese reducti$e theories aim to radically naturalise the notion of
representation! since they treat scientist(s purposes and intentions as non)essential to
representation% I distinguish &et*een the means and the constituents of representation!
and I argue that similarity and isomorphism are common &ut not uni$ersal means of
representation% I then present four other arguments to sho* that similarity and
isomorphism can not &e the constituents of scientific representation% I finish &y loo+ing at
the prospects for *ea+ened $ersions of these theories! and I argue that only those that
a&andon the aim to radically naturalise scientific representation are li+ely to &e
successful%
,
,% 'heories of Scientific Representation%
Many philosophers of science *ould agree that a primary aim of science is to
represent the *orld -Cart*right -.///0! 1iere -,233! .///0! 4riedman -,23.! chapter 5I0!
6itcher -,2370! Morrison -.//,! chapter II0! Morrison and Morgan -,2220! 5an 4raassen
-,23,! ,23809 a *ell +no*n dissenter is Ian :ac+ing -,23700% ;hat those philosophers
understand &y <represent= is ho*e$er a lot less clear% >o account of representation in
science is *ell)esta&lished% Perhaps this is not surprising% Consider the follo*ing four
$ery different e?amples of successful scientific representation! dra*n from engineering
and mathematical physics respecti$ely: a toy model of a &ridge9 an engineer(s plan for a
&ridge! such as the 4orth Rail @ridge A an e?ample carefully documented &y Michael
@a?andall -,23B! chapter ,09 the &illiard &all model of gases -:esse! ,2C809 and the
Duantum state diffusion eDuation for a particle su&Eect to a localiFation measurement
-Perci$al! ,2220% ;hat could there &e in common &et*een such disparate models that
allo*s them to representG
I choose these e?amples mainly &ecause they illustrate the range and $ariety of
representational de$ices in science% In these e?amples *e may usefully distinguish
&et*een the source and the target of the representation% Roughly! the source is the $ehicle
of the representation! the target is its o&Eect% In the first t*o e?amples the source is a
concrete physical o&Eect and so is the target% In the third e?ample *e may descri&e the
.
source as a physical system and the target as a state of a nature% In the fourth e?ample the
target is a physical phenomenon and the source a mathematical entity! an eDuation% In all
these cases A is the source and @ is the target *hen and only *hen <A represents @= is
true%
'here are of course many other +inds of representational media in science% 'he
sources of scientific representations may &e concrete physical o&Eects! systems! models!
diagrams! images or eDuations9 and similarly for possi&le targets% 'he only thing that they
ha$e in common is that they all putati$ely include real entities in the *orld9 and there
does not seem to &e any property in particular that allo*s any of them to perform one or
the other function%
I ta+e it that a su&stanti$e theory of scientific representation ought to pro$ide us *ith
necessary and sufficient conditions for a source to represent a target% It is natural to
e?pect these conditions to agree *ith our underlying intuitions a&out ordinary
representation in general9 &ut *e should not necessarily reDuire the conditions of
scientific representation to &e identical to those for ordinary representation% >either
should *e reDuire that a theory of scientific representation &e a&le to e?plain ho* humans
ha$e e$ol$ed the capacity to generate representations! or mental images of the *orld9
although this is an independently interesting issue -see e%g% ;oodfield! ,22,0
In addition! a good theory may pro$ide us *ith insight into some of the features that
are normally associated *ith scientific representations such as accuracy! relia&ility! truth!
7
empirical adeDuacy! e?planatory po*er9 &ut again *e shall not assume that this is a
reDuirement% In other *ords! *e shall not reDuire a theory of representation to mar+ or
e?plain the distinction &et*een accurate and inaccurate representation! or &et*een
relia&le and unrelia&le one! &ut merely &et*een something that is a representation and
something that isn(t%
,
'his presupposes a distinction &et*een the conditions for x to &e a
representation of y! and the conditions for x to &e an accurate or true representation of y%
@oth are important issues! &ut they must addressed and resol$ed separately% Science often
succeeds at constructing representations of phenomena! &ut it rarely succeeds at
constructing completely accurate ones -see e%g% @ailer)Hones! .//70% On disco$ering
particular inaccuracies in the representation *e are $ery rarely inclined to *ithdra* the
claim that it is a representation% 'hus a graph can &e a more or less accurate
representation of a &ridge! and a Duantum state diffusion eDuation can &e a more or less
accurate representation of a particular instance of localisation% I ha$e here little to say
a&out *hat ma+es one representation more accurate than another%
.
In this paper I critically discuss t*o proposals for a su&stanti$e theory of scientific
representation along these lines% 'he intuition underlying these theories may at first
appear natural and per$asi$e! &ut I *ill argue that on careful analysis it must &e resisted%
'he intuition is that a source A is a representation of a target @ if and only if A! or some
of its parts or properties! constitute a mirror image of @! or some of its parts or properties%
A and @ are entities occurring in the *orld as descri&ed &y science! so a thorough
scientific in$estigation of all the facts a&out A and @ and their relation should thus suffice
to settle the matter% 'his is perhaps &est summarised &y means of a slogan: <scientific
I
representation is a factual relation &et*een entities in the *orld that can &e studied &y
science=% Since the relation of representation is factual it can not in$ol$e essential or
irreduci&le Eudgements on the part of agents%
One sense in *hich *e may naturalise a concept is &y reducing it to facts! and thus
sho*ing ho* it does not in any essential *ay depend upon agent(s purposes or $alue
Eudgements -Putnam! .//.9 5an 4raassen! .//.0% 'he t*o theories that I criticise here are
naturalistic in this sense! since *hether or not representation o&tains depends on facts
a&out the *orld and does not in any *ay ans*er to the personal purposes! $ie*s or
interests of enDuirers% 'hese theories ha$e the $irtue of guaranteeing the o&Eecti$ity of
scientific representation *hich! unli+e linguistic representation perhaps! is certainly not a
matter of ar&itrary stipulation &y an agent%
:o*e$er! other non)naturalistic conceptions may guarantee the appropriate le$el of
o&Eecti$ity of scientific representations as *ell% In this paper I argue that the t*o main
naturalistic alternati$es are mista+en! thus pointing to the conclusion that no su&stanti$e
naturalistic theory of scientific representation *ill succeed% I am certainly not the first to
criticise similarity and isomorphism theories% 4or instance! Cummins -,232! chapter 70
discusses and reEects similarity theories of mental representation! and "o*nes -,22.0
criticises isomorphism as characterising the empirical adeDuacy of scientific theories% I
focus my critiDue on similarity and isomorphism as theories of scientific representation!
and I argue that &oth fail for precisely the same set of reasons%
B
.% Representation >aturalised: Similarity and Isomorphism%
;hat sort of factual relation must hold &et*een A and @ for A to represent @G 4or
instance! *hat relation must hold &et*een the graph of a &ridge! and the &ridge it
representsG It is o&$ious that not any ar&itrary relation &et*een A and @ *ill do: for there
are all sorts of relations that o&tain &et*een A -e%g% the graph0 and @ -e%g% the &ridge0!
*hich are irrele$ant to the representational relation itself A such as <&eing an artefact=! or
<&eing at least ,/ cm% long=% 'he success of the proEect of naturalising representation is
crucially dependent upon finding a suita&le type of relation that can fill in this role% 4or
the theory of representation to &e su&stanti$e in my sense it is reDuired that this relation
o&tains uni$ersally &et*een the source and the target! in all instances of successful
scientific representation%
'*o accounts ha$e &een a$aila&le in the literature for some time: similarity and
isomorphism% Ronald 1iere -,233! .///0 has defended the importance of similarity for
representation! *hich has also &een stressed for instance &y Aronson! :arrJ and ;ay
-,2270% @as $an 4raassen -,22,! ,22I0 has concentrated on the $irtues of isomorphism9
and other *riters in the structuralist tradition! including most prominently @rent Mundy
-,23C0! ha$e appealed to *ea+ened $ersions of isomorphism%
;e may enunciate the corresponding theories as follo*s:
7
C
'he similarity conception of representation KsimL: A represents @ if and only if A is
similar to @%
'he isomorphism conception of representation KisoL: A represents @ if and only if the
structure e?emplified &y A is isomorphic to the structure e?emplified &y @%
Similarity is a generalisation of resem&lance% '*o o&Eects resem&le each other if
there is a significant similarity &et*een their $isual appearance% KsimL does not assert that
resem&lance is a necessary and sufficient condition for representation9 it is a *ea+er
condition! *hich neither reDuires nor includes similarities in $isual appearance! or a
threshold <significant= amount of similarity% 'he follo*ing is typically assumed: A and @
are similar if and only if they share a su&set of their properties% In accordance *ith this
identity)&ased theory similarity is refle?i$e -A is ma?imally similar to itself0! and
symmetric -if A is similar to @! on account of sharing properties p,! p.! M pn! then @ is
similar to A on the same grounds09 &ut non)transiti$e -A may share p, *ith @! and @ may
share p. *ith C! *ithout A and C sharing any property A other than the property of sharing
a property *ith @N0%
Isomorphism is only *ell defined as a mathematical relation &et*een e?tensional
structures% :ence the a&o$e definition presupposes that any t*o o&Eects that stand in a
representational relation e?emplify isomorphic structures% 'he notion of structure)
e?emplification turns out to &e ridden *ith difficulties9 &ut the definition has the $irtue
that it ma+es sense of o&Eect)to)o&Eect representation outside pure mathematics% 'he
8
claim that t*o physical o&Eects A and @ are isomorphic is then short)hand for the claim
that the e?tensional structures that A and @ e?emplify are isomorphic% In *hat follo*s
<A= *ill indistinguisha&ly denote the source and the structure that it e?emplifies! and <@=
*ill denote the target and the structure that it e?emplifies% Isomorphism then demands
that there &e a one)to)one function that maps all the elements in the domain of one
structure onto the elements in the other structure(s domain and $ice)$ersa! *hile
preser$ing the relations defined in each structure% :ence A and @ must possess the same
cardinality% More precisely! suppose that A O P "! P
n
EQ and @ O P E! '
n
E Q9 *here "! E are
the domains of o&Eects in each structure and P
n
E and '
n
E are the n)place relations defined in
the structure% A and @ are isomorphic if and only if there is a one)to)one and onto
mapping f: " E! such that for any n)tuple -?,!M! ?n0 ": P
n
E K?,!M! ?nL only if '
n
E
Kf-?,0!M! f-?n0L9 and for any n)tuple -y,!M! yn0 E: '
n
E Ky,!M! ynL only if P
n
E Kf
),
-y,0!M! f
)
,
-yn0L% In other *ords! an isomorphism is a relation preser$ing mapping &et*een the
domains of t*o e?tensional structures! and its e?istence pro$es that the relational
frae!or" of the structures is the same%
I
It is possi&le in general to understand isomorphism as a form of similarity% 4or
suppose that A and @ are isomorphic9 then they share at least one property in common!
namely their relational frae!or"% :ence t*o isomorphic structures are similar! &ecause
their relational frame*or+s are identical% So t*o o&Eects that e?emplify isomorphic
structures are ipso facto structurally similar% 'he similarity in case -.0 &et*een the &ridge
and its graph is precisely of this type% 'his is prima facie an interesting ad$antage that
similarity enEoys o$er isomorphism% 4or neither similarity nor resem&lance can in general
3
&e reduced to isomorphism% Hudgements of similarity unpro&lematically apply to any sort
of o&Eects! including for instance perceptual e?periences! and it is unclear to say the least
ho* these e?periences could &e said to e?emplify structures at all% ;hether or not such
reduction is ultimately theoretically possi&le! in no ordinary conte?t are *e a&le to
translate Eudgements of similarities in! say! taste! to isomorphisms &et*een anything li+e
<taste structures= of different types of food% Analogously for most Eudgements of
resem&lance% 'he &asic pro&lem is that similarity and resem&lance are ordinarily and
unpro&lematically applied to &oth response)dependent and intensionally defined
properties! *hile isomorphism is not%
@ut *hat a&out those cases of representation *here the source and target can &e
ascri&ed an e?plicit structural e?emplificationG Argua&ly! many scientific representations
are of this sort% @ut e$en in these cases the reduction of similarity to isomorphism is
typically only possi&le conditional on the appropriate e?emplification of structure% '*o
o&Eects may &e similar in sharing Eust some of their properties! such as i%e% the colour
distri&ution of their surfaces% So only the structures defined &y the colour relation may &e
isomorphic% ;hile it is correct to claim that such o&Eects are similar! the isomorphism
claim must &e restricted to the specific properties shared%
#et us then suppose that either KisoL or KsimL *ere correct% It follo*s that to esta&lish
in cases ,)I that the source is a genuine representation of the target! *e need to
in$estigate the properties of the source and those of the target! and the relationship
&et*een them% >o further in$estigation is reDuired% Representation *ill o&tain if the right
2
type of relational facts o&tain &et*een A and @! independently of any agent(s Eudgements
on the matter% 'hus if *e can sho* KisoL or KsimL to &e correct *e *ill ipso facto ha$e
naturalised the notion of scientific representation% 'his! I thin+! is to a large e?tent the
moti$ation and dri$ing force &ehind the KisoL and KsimL conceptions%
4or e?ample! the aim to naturalise scientific representation is clear in 1iere(s *or+
and may in fact &e ta+en as a constant in his intellectual traEectory -see e%g% 1iere! ,233!
,222a! ,222&! forthcoming0% @ut *e must &e $ery careful to distinguish clearly the
different strands of naturalism present in his *or+% 'here are at least t*o clearly distinct
claims: there is a *ea+ form of naturalism that merely claims that science can study
representation9 and a stronger form of naturalism! *hich I employ in this paper! that
claims that the relation of representation does not in$ol$e in any essential *ay agent(s
intentions and $alue Eudgements! &ut appeals only to the facts% O$er the years 1iere has
mo$ed from a defence of &oth claims to a defence of the *ea+er claim only% So his recent
defence of naturalism -1iere! ,222a and forthcoming0 is compati&le *ith my reEection of
the naturalistic theories of representation discussed in this paper% -I am inde&ted to
illuminating discussions *ith Ron 1iere and @as 5an 4raassen on this point0%
7% Means and Constituents of Representation%
I *ant to first distinguish the means and the constituents of representation% In practice
the main purpose of representation is surrogati$e reasoning -S*oyer! ,22,0% Suppose! for
,/
instance! that an o&Eect A represents an o&Eect @9 then A must hold some particular
relationship to @ that allo*s us to infer some features of @ &y in$estigating A% 'a+e for
instance the e?ample of the phase space representation of the motion of a classical
particle% 'he graph may &e similar in respects a!&!c to the particle(s motion9 and *hen *e
reason a&out the graph in order to infer features of the particle(s motion *e do so &y
studying precisely that similarity% 'he means of the representation are thus those relations
&et*een A and @ that *e acti$ely ma+e use of in the process of inDuiring a&out @ &y
reasoning a&out A% >otice crucially that an o&Eect A or system may hold more than one
type of relation to another @! &ut at any one time only one of these *ill &e the means of
representation% 4or e?ample! a phase space graph of the motion of a paper &all in air may
&e &oth structurally isomorphic to the &all(s motion in space! and in addition similar to
the &all in &eing dra*n on the same type of paper% 'he similarity o&tains &ut is not the
means of the representation in this case -although there are circumstances in *hich it
could &e! for e?ample if *e *ere in$estigating the properties of paper! not motionN0
'hus there may &e a great $ariety of means &y *hich representation does its *or+:
isomorphism and similarity are Eust t*o common ones! &ut there are others! such as
e?emplification! instantiation! con$ention! truth% In addition! the means of representation
are not e?actly transparent: no source *ears its means of representation <on its slee$e=% In
many cases the actual means of a representation may &e opaDue to the uninitiated%
Consider a &u&&le cham&er photograph! an astronomical chart! or an eDuation of motion%
'o correctly understand *hat and ho* these sources ground inferences a&out their
representational targets in$aria&ly reDuires informed and s+ilful Eudgement% >ormally
,,
only one among the many relations o&taining &et*een A and @ is intended to pro$ide
grounds for such inferences% So much is common lore! particularly in the philosophy of
art% It is surprising therefore that the implications of this simple o&ser$ation regarding the
nature of scientific representation seem not to ha$e &een pic+ed out% In particular! I *ill
argue! it follo*s that neither KisoL nor KsimL! on their o*n! can account for the means of
scientific representation%
At this point the distinction &et*een the means of representation and its constituents
may &e dra*n as follo*s% 'he fact that *e use a particular relation -say! similarity0
&et*een A and @ to! say! infer @(s properties &y reasoning a&out A(s properties! should
not &e ta+en to mean that this relation is *hat constitutes the representation &y A of @%
'here could &e a deeper! hidden relation &et*een A and @% Suppose that A -for instance! a
phase space structure0 represents @ -the motion of a particle in space0 in $irtue of an
isomorphism% 'his appears to &e consistent *ith the fact that sometimes in reasoning
successfully a&out @ on the &asis of A *e need not employ or refer e?plicitly to the
isomorphism of A and @! &ut are a&le to use some other relationship instead% 4or instance!
on a particular occasion it may &e possi&le to in$estigate the properties of a particle(s
motion merely &y in$estigating its similarity -i%e% shared properties! such as for instance
the appearance of randomness0 *ith its phase space graph% It *ould appear then that in
this case the means of the representation -similarity0 fail to agree *ith its deeper
constituents -isomorphism0%
;e may then consider the follo*ing definitions:
,.
Means of Representation
At any time! the relation R &et*een A and @ is the means of the representation of @ &y A
if and only if! at that time! R is acti$ely considered in an inDuiry into the properties of @
&y reasoning a&out A%
Constituents of Representation
'he relation R &et*een A and @ is the constituents of the representation of @ &y A if and
only if R(s o&taining is necessary and sufficient for A to represent @%
'he distinction opens up a promising a$enue for defending KisoL and KsimL% One could
ta+e KsimL -KisoL0 as the &asic constituti$e notion at the heart of representation! *hich
*arrants the representational relation! *hile accepting that isomorphism -similarity0 may
&e employed as useful means once KsimL -KisoL0 has &een esta&lished% 4or e?ample! the
isomorphism &et*een a phase space structure and the motion of a particle could &e said
to &e the efficient means for applying the rele$ant similarities of structure that *arrant
KsimL! and hence representation% Or! alternati$ely! the o&ser$ed similarity &et*een t*o
&ridges may &e ta+en to &e merely an efficient means for us to ta+e cogniti$e ad$antage
of the deep structural isomorphism e?isting &et*een the &ridge(s structures: the similarity
is only a means to more efficient reasoning! &ut it is the isomorphism that is actually
*arranting the representational relation! in accordance *ith KisoL% :ence KsimL -KisoL0
,7
may fully characterise the constituents of the relation of representation! *hile failing to
characterise its means%
I% 4i$e arguments against similarity and isomorphism
I *ill no* present fi$e arguments against KsimL and KisoL% Some of these arguments
*ere first ad$anced &y SuRreF -,2220 and ha$e &een reiterated &y 4rigg -.//.0% 'he first
argument is the simple empirical o&ser$ation that neither KsimL nor KisoL can &e applied to
the full $ariety of uses of representational de$ices that crop up in the practice of science%
:ence an analysis of the means of representation in terms of Eust one of these conditions
*ould &e unduly restricti$e and local% :o*e$er! as I pointed out a&o$e! the defenders of
KisoL and KsimL ha$e an easy retreat: they can argue that KisoL and KsimL are meant as
su&stanti$e theories of the constituents! not the means of the representational relation9
they are meant to descri&e the relation that must o&tain &et*een A and @ for A to
represent @! independently of *hat relations are actually employed &y enDuirers in
dra*ing inferences a&out @ on the &asis of A% 'he retreat is perfectly honoura&le and
legitimate! for it is line *ith the pretensions of a su&stanti$e naturalised theory of
representation%
:o*e$er! four other arguments sho* that e$en in those cases *here KisoL and KsimL
apply! the analysis they yield is incorrect9 in other *ords! the isomorphism and similarity
conceptions can not on their o*n constitute representation% 'he second argument is that
,I
KisoL and KsimL lac+ some of the logical properties of representation% 'he third argument
is that they do not allo* for misrepresentation or inaccurate representation% 'he fourth
argument is that KsimL and KisoL are not necessary for representation A they fail in some
cases of successful representation% 'he fifth and final argument is that neither KisoL nor
KsimL can &e sufficient for representation! &ecause they lea$e out the essentially
directionality of representation%
I%,% 'he Argument from 5ariety: KsimL! KisoL do not apply to all representational de$ices%
Although similarity and isomorphism are among the most common means of
representation in science neither one! on its o*n! co$ers e$en nearly the *hole range% ;e
ha$e some firmed up intuitions! I thin+! a&out the means of representation in the four
cases mentioned:
Case , -'oy &ridge representation0: Similarity is almost al*ays the means for
concrete physical representations of concrete physical o&Eects% An engineer(s toy &ridge
may &e similar to the &ridge that it represents in the proportions and *eights of the
different parts! the relati$e strengths of the materials and the geometric shape% It is &y
reasoning on the &asis of these similarities that the source does its representational *or+%
'here are also important dissimilarities! such as siFe! *hich ma+e the representation only
a partially successful one! &ut similarity again seems to &e a good guide to determining
*hich parts are representational and *hich aren(t% @y contrast isomorphism! *hich is
,B
*ell)defined only as a relation &et*een mathematical structures! does not apply directly
to the relation &et*een t*o physical o&Eects descri&ed in -case ,0% @ut it does apply to
some a&stract structures that are e?emplified &y these t*o o&Eects! such as their geometric
shape%
:o*e$er! the representational use of the toy &ridge is almost al*ays grounded on
actual reasoning a&out its properties! along *ith those of the real &ridge! and not on the
properties of the structures e?emplified &y either &ridge% 'he means of the relation of
representation are not in this case captured &y the KisoL conception &ecause this
conception misidentifies its relata! *hich are the physical o&Eects themsel$es! and not the
structures e?emplified% 'o ma+e this point $i$id suppose! for instance! that t*o concrete
toy &ridges e?emplify e?actly the same geometric structure! isomorphic to that of the
larger &ridge% ;e typically treat these t*o &ridges as t*o different means and as distinct
representations of the same o&Eect! &ut an isomorphism analysis of the means of
representation does not allo* us to do that: the relationship R that each toy &ridge holds
to the larger &ridge is e?actly the same%
B
Case . -1raph of &ridge0: 'he range and depth of the dissimilarities &et*een the
source and the target &ecome greater in this case: a piece of paper containing the graph of
a &ridge is only similar to the &ridge it represents *ith respect to the geometric shape and
proportions &et*een the different points9 nothing else is interestingly similar% 'his
<similarity of structure= is &etter captured &y the alternati$e conception KisoL% Maps! plans
and graphs are typical cases *here isomorphism is the means of scientific representation%
,C
Case 7: -@illiard &all model0: 'his case appears to &e harder for &oth conceptions%
A system of &illiard &all is in not prima facie in any rele$ant sense similar to any state of
nature% ;e may any*ay refer to the relation of similarity or isomorphism that can o&tain
&et*een t*o to+en instances of these things% If so *e must ma+e sure that *e are
referring to a similarity &et*een the dynamical properties of the systems! collecti$ely
ta+en! as a system of &illiard &alls is similar to a system of gas molecules only in its
dynamical properties! and in no properties of the entities ta+en indi$idually at any one
time A other than their elasticity% Mutatis mutandis for isomorphism: this o&tains only
&et*een the mathematical structures e?emplified &y the dynamics of the systems! and not
&et*een the structures e?emplified &y the indi$idual entities%
Case I -Suantum state diffusion eDuation0: 'his case is simply not co$ered &y the
similarity analysis at all% A mathematical eDuation! *ritten do*n on a piece of paper!
represents a certain physical phenomenon &ut is not similar to it in any rele$ant respect% If
the eDuation is dynamical! one may focus on the phase)space structure defined &y the
eDuation! and on that structure *hich is &est e?emplified &y the phenomenon: if the
eDuation is an accurate representation of the phenomenon! isomorphism *ill o&tain
&et*een them and! as noted in section ,! isomorphism is a case of similarity%
@ut e$en KisoL is pro&lematic here% In most cases of mathematical representation it
seems farfetched to assume that the means of the representation is an isomorphism% It
tri$ially is the case that the dynamic phase space structure e?emplified &y a differential
,8
eDuation must &e isomorphic to the dynamic structure e?emplified &y the phenomenon! if
the eDuation accurately represents the phenomenon% @ut *hen scientists reason a&out a
differential eDuation in order to inDuire into the phenomenon it represents! they rarely
include an in$estigation of the formal properties of these structures% ;hat they acti$ely
do is loo+ for solutions to the eDuation gi$en certain &oundary conditions! and then chec+
*hether some parameters of those solutions correspond to o&ser$ed features of the
phenomena% 'he isomorphism *hich o&tains is not *hat they e?plicitly reason a&out! so
it is not in this case the means of the representation%
I%.% 'he #ogical Argument: KsimL! KisoL do not possess the logical properties of
representation
A su&stanti$e theory must ma+e clear that scientific representation is indeed a
type of representation9 i%e% that it shares the properties of ordinary representation%
Representation in general is an essentially non)symmetric phenomenon: a source is not
represented &y a target merely in $irtue of the fact that the source represents the target%
'here may &e conte?ts in *hich symmetry o&tains! &ut e$en there it is not automatic:
merely &ecause a photograph portrays a person! in some conte?t! it does not follo* that
the person stands for the photograph! in the same conte?t9 merely &ecause an eDuation
represents a phenomenon! the phenomenon can not &e said to stand for the eDuation%
Representation is also non)transiti$e and non)refle?i$e -as a matter of fact representation
,3
is pro&a&ly in addition irrefle?i$e! asymmetric and intransiti$e9 &ut I *ill not here need to
argue that much0% A theory of scientific representation must do Eustice to these features%
>elson 1oodman -,28B! pp% 7),/0 used these logical properties of representation to argue
against resem&lance theories! and his argument carries o$er against KsimL and KisoL% I shall
pursue here an illuminating analogy *ith painting A a particularly apt analogy in this
conte?t! as KisoL and KsimL &oth assume that scientific representation is essentially an
o&Eect)to)o&Eect relation rather than *ord)to)o&Eect! or mental state)to)o&Eect relation%
'hat is! &oth KisoL and KsimL assume that &oth relata of the relation of representation are
similarly structured entities endo*ed *ith properties% 'his e?plains *hy KisoL and KsimL
ha$e &een particularly attracti$e to defenders of the semantic $ie* of theories! since on
this $ie* theories are not linguistic entities &ut structures%
#

'he argument is! ho*e$er! independent of the analogy! and is in no *ay e?hausted
&y it% 'he purpose of the analogy is to call attention to the logical properties of o&Eect)to)
o&Eect representation in general! thus suggesting that scientific representation must
display these properties too% It could ho*e$er turn out that scientific representation is not
a +ind of o&Eect)to)o&Eect representation! or not entirely so9 &ut this is a possi&ility that
*ould ipso facto refute the KisoL and KsimL conceptions that I criticise here% 'o defeat the
argument! one *ould ha$e to sho* that KisoL and KsimL ha$e the logical properties of
representation in general! *hich I thin+ is patently not the case%
,2
Representation is non)refle?i$e: "iego 5elRFDueF(s portrait of Pope Innocent T
represents the Pope as he *as posing for 5elRFDueF &ut it does not need to represent the
portrait itself% Admittedly! 5elRFDueF astounded the *orld *ith the stri+ing &uilt)in
refle?i$ity of #as Meninas! *hich represents among other things! the act of its &eing
painted% A creati$e o&session *ith representing the $ery elusi$e act of representation is
part of art since at least the Suantrocentto% @ut e$en in these cases the representation
typically adds to the o&Eect! and also su&tracts from it: the source and the target are not
e?actly identical%
It *ould &e eDually *rong to claim that the Pope represents the painting% ;e may
put aside issues a&out *hether a non)e?isting o&Eect can &e said to represent: e$en *hen
the Pope *as sitting do*n posing for 5elRFDueF it *ould not ha$e &een right to claim that
he represented the painting% Representation is non)symmetric! at least! since it is often
one)*ay% Apparent cases of symmetry! such as some of Escher(s dra*ings! or t*o mirrors
placed opposite and reflecting each other! turn out to &e cases *here there is a distinct
representational relationship going each *ay%
'he recently deceased painter 4rancis @acon *as o&sessed *ith 5elRFDueF(s
portrait of Innocent T! and produced a large num&er of $ariations of his o*n! all of them
intending to represent the 5elRFDueF can$ass% 'he @acon portraits represent not the Pope!
&ut the 5elRFDueF can$ass% Or! suppose that a tourist ta+es a photograph of the room
*here #as Meninas hangs in the Prado9 that photograph represents the can$ass! not the
Spanish Royal 4amily% Representation is at least non)transiti$e! since apparent cases of
./
transiti$ity turn out on inspection to in$ol$e different representational relations &et*een
A and @ and &et*een @ and C%
:o*e$er! similarity is refle?i$e and symmetric! and isomorphism is refle?i$e!
symmetric and transiti$e% A glass of *ater is similar to itself! and similar to any other
glass that is similar to it% Mutatis mutandis for isomorphism: A geometric structure -a
sDuare0 is isomorphic to itself! and al*ays isomorphic to any other structure -another
sDuare of perhaps a different siFe0 that is isomorphic to it! and isomorphic to any structure
that is isomorphic to a structure that is isomorphic to it -an e$en larger sDuare0%
I%7% 'he Argument from Misrepresentation: KsimL! KisoL do not ma+e room for the
u&iDuitous phenomena of mistargetting andUor inaccuracy%
Misrepresentation is an u&iDuitous phenomenon in ordinary)life representation% It
comes in t*o $arieties% 'here is first the phenomenon of mista+ing the target of a
representation! or as I call it mistargetting: often *e mista+enly suppose the target of a
representation to &e something that it actually does not represent% Suppose that a friend of
mine has disguised himself to loo+ roughly similar to Pope Innocent T in the rele$ant
respects% In seeing the 5elRFDueF can$ass for the first time I am struc+ &y this
resem&lance and! ignoring the history and true target of the representation! I go on to
suppose that the 5elRFDueF represents my friend% 'his is a clear case of misrepresentation!
&ut there is no failure of similarity to e?plain it% Indeed misrepresentation &y accidental
similarity *ould &e impossi&le if KsimL *ere true! precisely &ecause similarity *ould then
.,
*arrant representation% E?actly the same argument goes mutatis mutandis for
isomorphism! and it is an argument that can &e easily transferred to cases of scientific
mistargetting% Consider the case of the Duantum state diffusion eDuation:
d| O )iUh : | dt V E -#E
W
#E ) X #E
W
#E A X #E
W
#E0 | dt V E -#E ) #E0 | dt%
'his eDuation represents the e$olution of the Duantum $ector state of a particle
su&Eect to a diffusion process% -'he first term is Eust the usual :amiltonian in the linear
SchrYdinger eDuation! the other t*o terms account for random diffusion and interaction
*ith a larger en$ironment%0 A mathematician *ho +ne* nothing a&out Duantum
mechanics *ould &e a&le to sol$e this eDuation for some &oundary conditions9 &y
accident the motion descri&ed may correspond to a particular classical particle(s
@ro*nian motion% 'his accidental fact on its o*n does not turn the eDuation into a
representation of the particle(s motion! ho*e$er! &ecause the essential directionality of
representation is missing%
'he point has &een argued persuasi$ely in the general case &y Putnam -,23,!
chapter ,0 and in the scientific case &y 5an 4raassen -,22I! p% ,8/0! and need not &e
rehearsed in detail here% It has long &een noted *ithin the philosophy of art too% 'hus
Richard ;ollheim *rites -,238! p% BI0:
<'he connection &et*een seeing)in and representation *as noted &y theorists of
representation &oth in antiDuity and in the Renaissance% Zet almost to a person
..
these thin+ers got the connection the *rong *ay round: they treated seeing)in as A
logically and historically A posterior to representation% 4or they held that!
*hene$er *e see! say! a horse in a cloud! or in a stained *all! or in a shado*! this
is &ecause there is a representation of a horse already there A a representation
made! of course! &y no human hand% 'hese representations! *hich *ould the *or+
of the gods or the result of chance! *ait for persons of e?ceptional sensiti$ity to
discern them! and then they deli$er themsel$es up%=
4or ;ollheim! li+e Putnam! 5an 4raassen and myself! the s+ill and acti$ity
reDuired to &ring a&out the e?perience of seeing)in -the appreciation &y an agent of the
[representational( Duality of a source0! is not a conseDuence of the relation of
representation &ut a condition for it%
Similarly for scientific representation% 4or suppose Putnam! 5an 4raassen and
;olheim *ere all *rong on this point:

A mathematician(s disco$ery of a certain ne*
mathematical structure -defined &y a ne* eDuation perhaps0 isomorphic to a particular
phenomenon *ould amount to the disco$ery of a representation of the phenomenon A
independently of *hether the mathematical structure is e$er actually applied &y anyone to
the phenomenon%

'he history te?t&oo+s *ould ha$e to &e re*ritten so that it *as
Riemann! not Einstein! *ho should get credit for first pro$iding a mathematical
representation of spacetime%
.7
'he second form of misrepresentation is the e$en more u&iDuitous! perhaps
uni$ersal! phenomenon of inaccuracy% Most representations are to some degree inaccurate
in some or other respects% KisoL can not account for inaccurate representation at all% 4or on
this conception either a model is a representation of! and thus isomorphic to! its target! or
it is not a representation at all% KsimL reDuires that the target and the source must share
some although not necessarily all their properties% :ence KsimL can account for the type
of inaccuracy that arises in an incomplete or idealised representation of a phenomenon!
i%e% one that lea$e out particularly salient features such as the highly idealised
representation of classical motion on a frictionless plane% @ut this *ill not al*ays help to
understand inaccurate representation in science! *here the inaccuracy is much more often
Duantitati$e than Dualitati$e% 4or e?ample >e*tonian mechanics! *ithout general
relati$istic corrections! can at &est pro$ide an appro?imately correct representation of the
solar system% Some motions *ould not &e Duite as predicted &y the theory! e$en if all
features of the solar system *ere to &e accounted for% 'he interesting Duestion is not *hat
properties fail to o&tain! &ut rather ho* far is the di$ergence &et*een the predictions and
the o&ser$ations regarding the $alues of the properties that do o&tain% KsimL offers no
guide on this issue%
I%I% 'he >on)>ecessity Argument: KsimL! KisoL are not necessary for representation A the
relation of representation may o&tain e$en if KsimL! KisoL fail%
It is tri$ial that any o&Eect is in principle similar to any other o&Eect% In fact the
point is often made that if all logically possi&le properties are permitted! then any o&Eect
.I
is similar to any other o&Eect in an infinite num&er of *ays! i%e% there is an infinite num&er
of properties that *e can concoct that *ill &e shared &et*een the o&Eects -<&eing on this
side of the moon=! <&eing neither &lac+ nor &lue=! etc0% If so! similarity *ould &e
necessary for representation &ut in a completely tri$ial *ay% 4or it *ould not only &e a
necessary condition on representation &ut also on non)representation%
'he defender of similarity may retort that it is not fair to include those logically
possi&le shared properties that ha$e nothing *hate$er to do *ith the representation itself
-such as <&eing on this side of the moon=0% A restriction is needed here to only those
properties or aspects of the source and the target that are <rele$ant= to the
representational relation: A represents @ if and only if A and @ are similar in the rele$ant
respects% It is not the case that any source is in principle tri$ially similar in the rele$ant
aspects to *hat it represents% Suppose that I am interested in representing in a painting the
colour of the ocean in front of me% I may represent the ocean &y painting some &lue and
green stripes on a piece of paper% Representation o&tains in this case if the colours on my
paper are similar to those of the ocean! and it fails other*ise: that is the only rele$ant
property% Any other logically possi&le similarity! such as <&eing on this side of the moon=!
is irrele$ant to this particular representation%
'here are t*o important o&Eections to this mo$e% 4irst! *hat is the criterion of
rele$ance in$o+ed hereG 'his criterion must presuma&ly lin+ rele$ance to the
representational relation itself! for other*ise there *ould &e no reason to e?pect rele$ant
similarity to &e necessary for representation% 'he shared properties that are rele$ant are
.B
precisely those that pertain to the representation% So! *e o&tain that! A represents @ if and
only if A and @ are similar in those respects in *hich A represents @% :o*e$er
illuminating this may &e a&out the actual use of similarity! it is circular as an analysis of
representationN
@ut then! and this is the second o&Eection! it isn(t straightfor*ard that similarity
*ith respect to rele$ant properties is in practice necessary for representation% 'his is
made most $i$id in the analogy *ith art! and to illustrate this point I li+e to in$o+e
1uernica! the *ell)+no*n painting &y Picasso% 'here are similarities &et*een parts of this
painting and many o&Eects! such as a &ull! a crying mother upholding a &a&y! an
enormous eye% 'he all seem undenia&ly rele$ant to the representational content of the
painting! if any similarities are! yet none of these similarities is a good guide to the actual
targets of the representation% 'here are at least t*o targets% Picasso *as interested in
representing the first e$er carpet)&om&ing of an entire ci$ilian population: the &om&ing!
under 4ranco(s consent if not direct orders -Preston! ,227! chapter IT0! of the @asDue
to*n of 1uernica &y :itler(s Condor #egion and Mussolini(s A$iaFione >aFionale in
,278% In addition 1uernica represents the threat of rising 4ascism in Europe! *hich is the
reason *hy it *as hugely effecti$e in &ringing *orld attention to the Spanish Repu&lic(s
cause% 'his is all historically *ell documented%
7
'he point is that none of the targets of
1uernica can &e easily placed in the rele$ant similarity relation *ith the painting! and
mutatis mutandis for isomorphism%
$
.C
'he case in science is not significantly different% An eDuation A i%e% the actual
physical signs on the paper\ is as dissimilar as it could &e from the phenomenon that it
represents% Mutatis mutandis for isomorphism! as *e ha$e already seen in the case of
inaccurate representation% ;e are perfectly happy *ith the claim that >e*tonian
mechanics pro$ides a representation of the solar system! e$en if it is clear that >e*tonian
mechanics! *ithout general relati$istic corrections! is empirically inadeDuate and non)
isomorphic to the phenomena of planetary motion% A possi&le retort on &ehalf of KisoL and
KsimL is that *e should concentrate entirely upon the su&set of properties! or the
su&structure! that corresponds to those motions that are correctly predicted% @ut in cases
of Duantitati$e inaccuracy this normally *on(t help% >e*tonian mechanics argua&ly does
not descri&e any actual planetary motion in a Duantitati$ely accurate *ay%
I%B% 'he >on)Sufficiency Argument: KsimL! KisoL are not sufficient for representation Athe
relation of representation may fail to o&tain e$en if KsimL! KisoL hold%
'he pre$ious four arguments already point to a feature of representation that is not
captured &y the KisoL or KsimL analyses: the essential directionality of representation% 'his
*as perhaps most apparent in the argument from misrepresentation: 'he o&Eect that
constitutes the source of a model has no directionality per se! &ut in a genuine
representational relation the source leads to the target% >either similarity nor isomorphism
can capture this capacity of the representational relation to ta+e an informed and
competent inDuirer from consideration of the source to the target% @ut it is this feature that
lies at the heart of the phenomenological non)symmetry of representation% Consider for
.8
instance t*o identical glasses% 'hey share all their -monadic0 properties! and are hence as
similar as they could &e% @ut neither of them leads to the other unless they are in a
representational relation! and then only that *hich is the source *ill ha$e the capacity to
lead to the target% Or consider the traEectory in phase space descri&ed &y the state $ector
of a Duantum particle% Un&e+no*nst to us this traEectory may *ell &e isomorphic to the
motion in physical space of a real classical particle% @ut unless the phase space model is
intended for the particle(s motion! the representational relation *ill fail to o&tain% :ence
neither similarity nor isomorphism are sufficient for representation%
'here is an additional reason *hy isomorphism is not sufficient for scientific
representation% It is related to the notion of structure)e?emplification% 1oodman -,28B!
pp% B. ff%0 pro$ided an analysis of the notion of e?emplification as a special class of
representation: If ? e?emplifies y then ? denotes y and y denotes ?9 &ut ? may denote y
*ithout e?emplifying it )) e?emplification reDuires denotation &oth *ays% My s*eater
e?emplifies red if and only if it &oth denotes red and is denoted &y red -i%e% the s*eater is
used to refer to red and also is red%0
>o* let us suppose that this analysis of e?emplification goes through for
structural representation% 'hen *hene$er o&Eect ? e?emplifies structure y it &oth
represents y and is represented &y y% It follo*s then that for an o&Eect A to represent some
o&Eect @ &y means of KisoL! the structure e?emplified &y A must &e isomorphic to the
structure e?emplified &y @% @ut that Eust means! if the supposition is right! that for A to
represent @ there must &e a structure that represents A isomorphic to a structure that
.3
represents @% And *e *ill no* *ant to as+ ho* the structures represent the o&Eects in the
first place%
4or instance! the Duantum state diffusion eDuation for a localising particle descri&es a
random *al+ motion in a phase)space structure% 'his structure represents not the
particle(s motion! &ut a representation of it! namely the motion of the $ector in :il&ert
space that corresponds to the state of the particle% @ut representing a representation of x is
in no *ay eDui$alent to representing x9 and *e are left *ith the Duestion of ho* x is
mathematically represented in the first place% So isomorphism is not in general sufficient
for representation% Perhaps parado?ically the case of representation of a *ell esta&lished
physical phenomenon &y means of a differential mathematical eDuation is the hardest
case for KisoL to accommodate% ->ote that I am not claiming that isomorphism is
irrele$ant to mathematical modes of representation in science! only that it does not
constitute representation9 it may for instance &e $ery useful in esta&lishing the accuracy
of a representation%0
B% 'he Amended 5ersions 4are no @etter
A recurrent theme in these arguments! *hich &ecame e?plicit in the discussion of the
non)sufficiency argument! is the appeal to the essential directionality of representation: a
necessary condition for A to represent @ is that consideration of A leads an informed and
competent inDuirer to consider @% I *ill refer to A(s capacity to lead a competent and
.2
informed enDuirer to consider @ as the representational force of A% Representational
forces are relational properties of sources in particular conte?ts of inDuiry% 'hey are
determined at least in part &y correct intended uses! *hich in turn are typically
conditioned and maintained &y socially enforced con$entions and practices: A can ha$e
no representational force unless it stands in a representing relation to @9 and it can not
stand in such a relation unless it is intended as a representation of @ &y some suita&ly
competent and informed inDuirer%
>ote that I am careful to refer to the essential directionality! not intentionality! of
representation% 'here has &een an important de&ate in the philosophy of mind in recent
decades that assumes that a source(s representational force is nothing &ut the
intentionality of the mental state of an agent that employs the source to represent some
target% 'his assumption is friendly to my critical analysis! since intentionality has the
right properties to ground the arguments that I propose in this paper against similarity and
isomorphism% 'here is no dou&t that the connection &et*een representational force and
intentional states reDuires further analysis9 ho*e$er! I neither need nor *ish to ma+e this
assumption here% At &est! gi$en the present lac+ of consensus a&out *hat intentionality
may &e! this assumption *ould &e a+in to trying to e?plain a child(s a&ility to ride a
&icycle &y appealing to his sense of &alance: e$en if no&ody dou&ts that there is a
connection! it is not $ery e?planatory since *e do not ha$e a clear understanding of ho*
children de$elop their sense of &alance% Analogously! it can not &e sound methodology to
in$o+e a difficult and o&scure notion -intentionality0 in order to e?plain a difficult &ut not
particularly o&scure human acti$ity -representation0% 'hat is *hy! at this stage at least! I
7/
prefer a plain intended)use theory of representational force that lea$es room for further
specification! *hich may or not include intentionality%
%
Can KisoL and KsimL &e made to *or+ &y simply amending them to account for this
directional componentG 'he amended $ersions *ould loo+ as follo*s:
KsimL(: A represents @ if and only if i0 A is similar to @ and ii0 the representational force
of A points to @%
KisoL(: A represents @ if and only if i0 the structure e?emplified &y A is isomorphic to the
structure e?emplified &y @ and ii0 the representational force of A points to @%
'he first thing to notice a&out these amended $ersions is that they a&andon the
aim of naturalising representation% Representation can no longer &e esta&lished &y means
of a scientific in$estigation of the facts of the matter A for there are elements in the
relation of representation! namely the representational forces in part ii0! that essentially
in$ol$e $alue Eudgements! and are not reduci&le to facts%
10
@ut in fact! KsimL( and KisoL( can not &e correct% Certainly the additional clause
stipulating the correct intended use of the representation turns conditions KisoL( and KsimL(
into sufficient conditions for representation! and the non)sufficiency argument no longer
applies% "epending on ho* *e e?plicate intended use! the logical argument might also
lose its force% @ut the other arguments still apply% 'he non)necessity argument is! if
7,
anything! strengthened as the necessary conditions on representation are no* stronger%
'he argument from $ariety sho*s that neither KisoL( nor KsimL( can descri&e all the means
of representation9 *hile the misrepresentation and non)necessity arguments sho* that
they do not pro$ide a su&stantial theory of the constituents of representation% Simply
adding further conditions to KisoL or KsimL to ma+e room for the essential directionality of
representation *ill not help%
C% ;ea+ening similarity and isomorphism%
'he prospects for a su&stantial naturalistic theory of representation seem &lea+%
Certainly KsimL and KisoL are non)starters% In this final section I ta+e a loo+ at a num&er of
attempts to *ea+en the conditions on representation imposed &y KisoL and KsimL% 'hese
programmes are either &eing tentati$ely de$eloped at present or could &e de$eloped% So
my conclusions ha$e a correspondingly tentati$e and pro$isional character%
C%,% Similarity *ithout Identity
'he pro&lem then lies not *ith *hat KisoL and KsimL lac+ &ut *ith *hat they($e
got% ;e must try to su&tract from! not add to! these conditions% One assumption that *as
&uilt into KsimL is the identity)&ased theory of similarity% 'his theory seems natural! gi$es
a high le$el of precision to the concept! and ma+es it possi&le for us to Duantify and
7.
measure degrees of similarity &et*een o&Eects -as ratios of properties shared0% @ut it may
&e mista+en%
E$idence has &een reported -&y Eileen ;ay! for e?ample! in a tal+ at the #as
Cruces modelling conference! >e* Me?ico! Hanuary .//.0 *ithin e?perimental *or+ in
cogniti$e psychology for a non)identity &ased understanding of similarity! *hich
emphasises the essential role of conte?tual factors and agent)dri$en purposes in similarity
Eudgements% -In the terminology of this paper! this turns KsimL into a non)naturalistic
theory%0 #et us suppose that similarity &et*een t*o o&Eects is not simply a case of sharing
a property! &ut a more comple? conte?tual relation% ;e don(t ha$e a $ery good
understanding of *hat this relation may &e! &ut Eileen ;ay argues that on such theories
of similarity there is typically no reason to e?pect similarity Eudgements to &e symmetric:
'he fact that A is similar to @ does not ipso facto reDuire @ to &e similar to A% If ;ay is
right the logical argument does not cut as strongly against similarity as it seemed% @ut it
applies nonetheless% 4or ho*e$er similarity is concei$ed! it must &e refle?i$e% If
something is not similar to itself then it is not similar to anything else% Any theory of
similarity must concede this: similarity comprises identity9 identity is a limiting case of
similarity% :ere representation and similarity definitely depart! for the $ast maEority of
representations patently do not represent themsel$es%
:o*e$er the com&ination of 1iere(s emphasis on the essentially pragmatic
character of similarity Eudgements -1iere! ,2330 *ith ;ay(s non)identity &ased
understanding of similarity *ould undenia&ly &ring similarity and representation closer%
77
And indeed 1iere has recently proposed to understand representation as a four)place
acti&ity% As he *rites: <'he acti$ity of representing! if thought as a representation at all!
should ha$e at least four places *ith roughly the follo*ing form: [S uses M to represent
; for purposes P( -1iere! forthcoming! p% TT0% 'his theory *ould &e successful to the
e?tent that it &uilds the source(s representational force! *ith all its normati$e import! into
the relation of representation itself% 'hat *ould turn it into a non)naturalistic theory in the
terminology of this paper% 'he non)sufficiency argument *ould ha$e no force against
such a theory! and neither *ould the mistargetting part of the argument from
misrepresentation! or the non)symmetry part of the logical argument% Refle?i$ity and the
non)necessity argument *ould remain the standing &loc+s for this interesting non)
naturalistic theory%
C%.% :omomorphism
EliFa&eth #loyd! a prominent defender of the semantic $ie*! suggests that <in
practice the relationship &et*een theoretical and empirical model is typically *ea+er than
isomorphism! usually a homomorphism! or sometimes e$en a *ea+er type of morphism=%
-#loyd! ,233! chapter .! footnote .0% Although #loyd is not in this passage specifically
referring to representation! the KisoL condition does get *ea+ened in a $ariety of *ays!
*hich sol$e some &ut not all of the pro&lems that I ha$e raised% 4or instance! follo*ing
the pioneering *or+ of 6rantF! Suppes et al -,28,0! @rent Mundy employs the notion of
homomorphism! and sho*s ho* to apply it to measurement theory! spacetime geometry
and classical +inematics% ;e say that an e?tensional structure A is faithfully
7I
homomorphic to an e?tensional structure @ if and only if there is a function that maps all
the elements in the domain of A into the elements in @(s domain! *hile preser$ing the
relations defined in A(s structure% More precisely! suppose that A and @ uniDuely
e?emplify the structures P "! P
n
EQ and P E! '
n
E Q9 *here "! E are the domains of o&Eects
in each structure and P
n
E and '
n
E are the n)place relations defined in the structure% 'hen A
is faithfully homomorphic to @ -Mundy! ,23C! p% 72B0 iff there is a mapping f: " E!
such that for any n)tuple -?,!M! ?n0 ": P
n
E K?,!M! ?nL if and only if '
n
E Kf-?,0!M! f-?n0L%
'he correspondingly *ea+ened $ersion of KisoL is:
'he homomorphism conception of representation KhomoL: A represents @ if and only if
the structure e?emplified &y @ is homomorphic to the structure e?emplified &y A%
An homomorphism is! unli+e an isomorphism! neither one)to)one nor onto! so the
cardinality of A and @ may differ% 'his feature *as notoriously used &y 6rantF! Suppes et
al -,28,0 to sho* that KhomoL rather than KisoL is appropriate for theories of
measurement% 'he important ad$antage that KhomL enEoys o$er KisoL is then the a&ility to
deal *ith partially accurate models% Parts of a source may not represent any of the aspects
of the homomorphic target% So the hope is that KhomL *ill &e a&le to refute the part of the
argument from misrepresentation that refers to inaccurate representation! and its
conseDuences for the non)necessity argument% 'he solar system may only &e represented
&y the part of the >e*tonian model that asserts the num&er of planets and their a$erage
pro?imity to the sun! *ithout specifying their precise motions% 'he highly de$eloped
structural theory of measurement as homomorphism into the real num&er continuum
7B
allo*s KhomoL to pro$ide precise estimates for these num&ers -6rantF el at! ,28,9 see
also "]eF! ,2280% It seems clear that the mo$e to KhomoL *ea+ens the non)necessity
argument -although interestingly it does not dispel the force of the art analogy in that
argument0%
:o*e$er! all the other arguments apply against KhomoL too% 'his includes the
argument from $ariety9 the mistargetting part of the argument from misrepresentation9
and the non)sufficiency argument% 'he logical argument is significantly *ea+ened &ut not
a$oided: :omomorphism is neither symmetric nor transiti$e! &ut it is refle?i$e%
C%7% Partial Isomorphism
Another proposal to *ea+en KisoL may &e pro$ided &y Mi+en&erg! "a Costa and
ChuaDui(s -,23C0 notion of partial structure! and the corresponding notion of partial
isomorphism introduced &y @ueno -,2280% A partial structure "! Ri,! Ri.! Ri7 defines for
each relation Ri a set of n)tuples that satisfy Ri! a set of n)tuples that do not satisfy Ri! and
a set of n)tuples for *hich it is not defined *hether they satisfy Ri or not% 1i$en t*o
partial structures A O "! Ri,! Ri.! Ri7 and @ O E! R(i,! R(i.! R(i7 <the function f: " E
is a partial isomorphism if i0 f is &iEecti$e! and ii0 for e$ery ? and y "! Ri,-?!y0 if and
only if R(i, -f-?0! f-y00 and Ri. -?!y0 if and only if R(i. -f-?0!f-y00= -@ueno! ,228! p% B2C9
4rench and #adyman! ,222! p% ,/30% 'he corresponding theory of representation *ould
then &e:
7C
'he partial isomorphism conception of representation Kpartial isoL: A represents @ if and
only if the structure e?emplified &y A is partially isomorphic to the structure e?emplified
&y @%
'he ad$ocates of partial isomorphism argue that the introduction of Ri7 ser$es to
accommodate the partiality and openness of the acti$ity of model &uilding% 'hat may &e
so! &ut as a theory of representation Kpartial isoL fares e$en *orse than KhomoL% Since
according to i0 f is &iEecti$e! it follo*s from ii0 that 'i3 (x,y) if and only if '*i3 (f(x),f(y))!
and hence partial isomorphism reduces to three separate isomorphisms% So it remains to
&e seen *hether Kpartial isoL can a$oid the inaccuracy part of the argument from
misrepresentation! and correspondingly *ea+en the non)necessity argument% E$en if this
could &e done! Kpartial isoL *ould &e at a disad$antage *ith respect to KhomoL since the
logical argument *eights e$en more strongly against Kpartial isoL: partial isomorphism!
unli+e homomorphism! is symmetric%
I%I% Structural Representation *ithout Isomorphism
Other *riters *ithin the structuralist tradition ha$e &een more cautious% It does
not follo* from the claim that theories -or models0 are! or contain! structures that the
relation that constitutes representation is a structural one% 'he arguments that I ha$e
presented in this paper suggest that *e should loo+ else*here for the constituents of
78
representation! perhaps e$en in those cases *here the source and the target of the
representation are structures%
Chris S*oyer -S*oyer! ,232! p% IB.9 see also "]eF! ,2230 for instance rightly
claims that <structural representation is not a necessary condition for representation in the
ordinary sense of the *ord! since *ith sufficient perser$erance A or per$ersity A *e can
use anything to represent $irtually anything else! and in many cases the t*o things *on(t
ha$e any interesting structural similarities at all% And it is not sufficient for ordinary
representation! since if you can find one structural representation of something! you can
usually find many%= S*oyer is also precisely right in characterising structural
representation as ha$ing the <potential= to &e used in surrogati$e reasoning a&out its
target%
After ha$ing considered si? different phenomenological constraints upon
structural representation! S*oyer proposes the notion of an U)morphism% -S*oyer(s
constraints implicitly rule out isomorphism! homomorphism and partial isomorphism as
the relation of structural representation! thus adding grist to the mill of my critiDue of
KisoL and its cousins0% Consider the representation of some structure @ &y means of
another structure A9 and consider t*o su&sets of @(s domain and % 'hen S*oyer(s
notion is as follo*s:
,,
A structure A structurally represents another structure @ iff there is
a -neither necessarily one)to)one nor onto0 mapping c+ , A that preser$es all the
relations defined o$er and counter)preser$es all the relations defined o$er ! *here
is non)empty% Since is non)empty! structural representation ser$es al*ays to carry out
73
surrogate reasoning a&out its target% S*oyer(s notion does not meet the logical!
misrepresentation and non)sufficiency arguments presented here -in particular U)
morphisms are refle?i$e09 neither is it meant to do so! since it is not meant as a theory of
scientific representation in general% Zet S*oyer(s *or+ sho*s that KisoL! KhomL and Kpart
isoL do not correctly descri&e e$en the means of structural representationN
8% Conclusions%
One type of naturalism urges us to reduce the relation of scientific representation to facts
a&out its relata% I ha$e argued that no theory that attempts to reduce scientific
representation to similarity or isomorphism *ill succeed% 'his might lead us to Duestion
the naturalism that ga$e rise to the need for reduction in the first place% 'here are other
types of naturalism! *hich may offer different strategies to successfully naturalise
scientific representation ) &ut they *ill certainly not in$ol$e a reduction of representation
to similarity or isomorphism%
72
Ac+no*ledgements:
'his paper is part of a fi$e)year long proEect! and it *ould carry a long list of
indi$idual ac+no*ledgements% @as 5an 4raassen! Ronald 1iere and t*o anonymous
referees of this Eournal stand out for useful comments and suggestions9 and Arthur 4ine
stands out for his encouragement and confidence in the proEect from its &eginning% I
*ould li+e to than+ all those *ho offered helpful comments and suggestions *hen I
deli$ered parts of the paper at >orth*estern Uni$ersity -,2230! Uni$ersities of Chicago
-,2230! @radford -.///0! #eeds -.///0! E?eter -.//.0! California at San "iego -.//. and
.//70! "amstaadt 'echnical Uni$ersity -.//.0! Santiago de Compostela -.//.0!
@arcelona -.//709 and at the International Conferences at Pa$ia -,2230! >e* Me?ico
State Uni$ersity -.//, and .//.0! "u&ro$ni+ -.//.09 the Italian Society for Analytical
Philosophy conference in @ergamo -.//.0! and the Philosophy of Science Association
&iennial conference in Mil*au+ee -.//.0% 'han+s to my students at the III Summer
School in 'heory of 6no*ledge! Madralin -.//,0! and to its organisers! particularly
RysFard ;oEcic+i% 'his *or+ has &een funded &y a #e$erhulme 'rust fello*ship -.//,)
.0! and &y the Spanish Ministry of Science and 'echnology proEects @44).//.)/,BB.
and @44 .//.)/,.II%
I/
References:
AR>:EIM! R% -,2C.0 Picasso(s 1uernica: 'he 1enesis of a Painting -@er+eley!
Uni$ersity of California Press0%
ARO>SO>! H%! :ARR^! R% _ ;AZ! E% -,2270 Realism Rescued -#ondon! "uc+*orth0%
@AI#ER)HO>ES! "% -.//70! ;hen scientific models represent! International Studies in
the Philosophy of Science! ,8%
@ATA>"A##! M% -,23B0 Patterns of Intention: On the :istorical E?planation of
Pictures ->e* :a$en! Zale Uni$ersity Press0%
@#U>'! A% -,2C20 Picasso(s 1uernica! -O?ford! O?ford Uni$ersity Press0%
@UE>O! O% -,2280 Empirical adeDuacy: a partial structures approach=! Studies in
:istory and Philosophy of Science .3: pp% B3B)C,/%
CAR';RI1:'! >% -,2220 'he "appled ;orld -Cam&ridge! Cam&ridge Uni$ersity
Press0%
C:IPP! :%@% -,2320 Picasso(s 1uernica! -#ondon! 'hames and :udson0%
I,
CUMMI>S! R% -,2320 Meaning and Mental Representation -Cam&ridge! Mass%! MI'
Press0%
CUMMI>S! R% -,2280 Representations! 'argets and Attitudes -Cam&ridge! Mass%! MI'
Press0%
CUMMI>S! R% -.///0 Reply to Milli+an! Philosophy and Phenomenological Research!
$ol% #T! >o% ,! pp% ,,7),.8%
"IE! H%A% -,2280 A hundred years of num&ers: an historical introduction to
measurement theory ,338),22/% Part I: the formation period! Studies in :istory and
Philosophy of Science! .3 U,! pp% ,C8),3B%
"IE! H%A% -,2280 A hundred years of num&ers: an historical introduction to
measurement theory ,338),22/% Part II: Suppes and the mature theory! Studies in :istory
and Philosophy of Science! .3 U.! pp% .78).CB%
"IE! H%A% -,2230 :acia una teor]a general de la representaci`n cient]fica! 'heoria! 5ol%
,7U,! pp% ,,7),72%
I.
"O;>ES! S% -,22.0 'he importance of models in scientific theoriFing: a deflationary
semantic approach! in: "% :U##! M% 4OR@ES _ 6% O6RU:#I6 -Eds%0 Proceedings of
the Philosophy of Science Association! $ol% ,! pp% ,I.),B7%
E1A>! 4% -,2230 Re$ie* of Cummins! R: Representations! 'argets and Attitudes! 'he
Philosophical Re$ie*! $ol% ,/8! no% ,! pp% ,,3),./%
E#1I>! C% -,2280 @et*een the A&solute and the Ar&itrary -Ithaca! Cornell Uni$ersity
Press0%
4RIE"MA>! M% -,23.0 4oundations of Space 'ime 'heories -Princeton! Princeton
Uni$ersity Press0%
4RE>C:! S% -forthcoming0 A model)theoretic account of scientific representation!
Proceedings of the Philosophy of Science Association! PSA meeting! Mil*au+ee!
>o$em&er .//.%
4RI11! R% -.//.0 Models and representation: *hy structures are not enough=! CP>SS
"iscussion paper series! "P MEAS .BU/.%
1IERE! R% -,2330 E?plaining Science: A Cogniti$e Approach -Chicago! Uni$ersity of
Chicago Press0%
I7
1IERE! R% -,222a0 Science *ithout #a*s -Chicago! Uni$ersity of Chicago Press0%
1IERE! R% -,222&0 Using models to represent reality! in: #% MA1>A>I!
>%>ERSESSIA> _ P% ':A1AR" -Eds%0 Model)@ased Reasoning in Scientific
"isco$ery ->e* Zor+! 6lu*er Academic Press0%
1IERE! R% -forthcoming0 <:o* models are used to represent physical reality=!
Philosophy of Science Proceedings! PSA meeting! Mil*au+ee! >o$em&er .//.%
1OO"MA>! >% -,28C0 #anguages of Art -Indianapolis! :ac+ett0%
1OO"MA>! >% -,2830 ;ays of ;orldma+ing -Indianapolis! :ac+ett0%
:AC6I>1! I% -,2370 Representing and Inter$ening -O?ford! O?ford Uni$ersity Press0%
:ESSE! M% -,2C80 Models and Analogies in Science ->otre "ame! Uni$ersity of Indiana
Press0%
6I'C:ER! P% -,2370 'he Ad$ancement of Science -O?ford! O?ford Uni$ersity Press0%
6RA>'! "%! #UCE! "%! SUPPES! P% _ '5ERS6Z! A% -,28,0 4oundations of
Measurement , ->e* Zor+! >e* Zor+ Academic Press0%
II
##OZ"! E% -,2330 'he Structure and Confirmation of E$olutionary 'heory ->e* Zor+!
1reen*ood Press09 reprinted &y -Princenton: Princeton Uni$ersity Press! ,22I0%
MAC"O>A#"! 1% -,2230 Re$ie* of Cummins! R: Representations! 'argets and
Attitudes! @ritish Hournal for the Philosophy of Science! $ol% I2! pp% ,8B),3/%
MI6E>@ER1! I%! "A COS'A! >% _ C:UASUI! R% -,23C0 Pragmatic truth and
appro?imation to truth=! Hournal of Sym&olic #ogic! B,! pp% ./,)..,%
MI##I6A>! R% -,23I0 #anguage! 'hought and Other @iological Categories -Cam&ridge!
Mass%! MI' Press0%
MI##I6A>! R% -.///0 Re$ie* of Cummins! R: Representations! targets and attitudes!
Philosophy and Phenomenological Research! $ol% #T! >o% ,! pp% ,/7),,7%
MORRISO>! M% -.//,0 Unifying Scientific 'heories -Cam&ridge! Cam&ridge Uni$ersity
Press0%
MORRISO>! M _ MOR1A>! M% -,2220 Models as autonomous agents! in:
MORRISO>! M% _ MOR1A>! M -Eds%0 Models as Mediators -Cam&ridge! Cam&ridge
Uni$ersity Press0%
IB
MU>"Z! @% -,23C0 On the general theory of meaningful representation! Synthese! C8!
pp% 72,)I78%
PE'ERSO>! E% -.//.0 <'he mediating models approach $ersus the semantic $ie* of
theories: 'o*ard a rapprochement=! Ph" dissertation! Uni$ersity of California San
"iego%
PERCI5A#! I% -,2220 Suantum State "iffusion -Cam&ridge! Cam&ridge Uni$ersity
Press0%
PRES'O>! P% -,2270 4ranco -#ondon! 4ontana Press0%
PU'>AM! :% -,23,0 Reason! 'ruth and :istory -Cam&ridge! Cam&ridge Uni$ersity
Press0%
PU'>AM! :% -.//.0 'he Collapse of the 4act)5alue "istinction -Cam&ridge! Mass%!
:ar$ard Uni$ersity Press0%
ROR'Z! R% -,23/0 Philosophy and the Mirror of >ature -Princeton! Princeton Uni$ersity
Press0%
IC
SUARE! M% -,2220 'heories! models and representations! in: #% MA1>A>I! >%
>ERSESSIA> _ P% ':A1AR" -Eds%0! Model)@ased Reasoning in Scientific "isco$ery!
->e* Zor+! 6lu*er Academic Press0! pp% 8B)37%
SUARE! M% -forthcoming0 An inferential conception of scientific representation!
Proceedings of the Philosophy of Science Association! PSA meeting! Mil*au+ee!
>o$em&er .//.%
S;OZER! C% -,22,0 Structural representation and surrogati$e reasoning! Synthese! 38!
pp% II2)B/3%
5A> 4RAASSE>! @% -,23,0 'he Scientific Image! -O?ford! O?ford Uni$ersity Press0%
5A> 4RAASSE>! @% -,2380 'he semantic approach to scientific theories! in: >%
>ERSESSIA> -Ed%0 'he Process of Science -"ordrecht! 6lu*er0! pp% ,/B),.I%
5A> 4RAASSE>! @% -,22.0 4rom $icious circle to infinite regress! and &ac+ again! in:
"% :U##! M% 4OR@ES A>" 6% O6RU:#I6 -Eds%0 Proceedings of the Philosophy of
Science Association! $ol% ,! pp% C).2%
5A> 4RAASSE>! @% -,22I0 Interpretation of science: Science as interpretation! in: H%
:I#1E5OOR" -Ed%0 Physics and Our 5ie* of the ;orld! -Cam&ridge! Cam&ridge
Uni$ersity Press0%
I8
5A> 4RAASSE>! @% -.///0 'he theory of tragedy and of science: does nature ha$e
narrati$e structureG! in: S4E>"O>I)ME>"OU! O% -Ed%0 Aristotle and Contemporary
Science! $ol% ,! ->e* Zor+! Peter #ang0%
5A> 4RAASSE>! @% -.//.0 'he Empirical Stance! ->e* :a$en! Zale Uni$ersity Press0%
;OO"4IE#"! A% -,22,0 'he emergence of natural representations=! Philosophical
'opics! $ol% ,3%.! pp% ,38).,7%
;O##:EIM! R% -,2380 Painting as an Art -#ondon! 'hames and :udson0%
I3
>ote on Contri&utor:
Mauricio SuRreF is Profesor 'itular in the "epartment of #ogic and Philosophy of
Science! Uni$ersidad Complutense de Madrid% :e gained his Ph" from the #ondon
School of Economics in ,228! *ith a thesis on the philosophical implications of Duantum
mechanics% :e has held a postdoctoral research fello*ship at >orth*estern Uni$ersity!
and has taught at #ondon! O?ford! St Andre*s and @ristol Uni$ersities%
Correspondence:
Prof% Mauricio SuRreF!
"epartamento de #`gica y 4ilosof]a de la Ciencia!
4acultad de 4ilosof]a <@=!
Uni$ersidad Complutense de Madrid!
.3/I/ Madrid! Spain%
Email: suare-.filos.uc.es
I2
,
>otes:
'his does not diminish the strength of my argument% On the contrary! since I *ant to argue that there
can &e no radically naturalistic and su&stanti$e theory of the constituents of scientific representation!
my argument *ould &e *ea+er and less interesting the stronger the independent conditions on
representation%
.
'he inferential conception that I present in -SuRreF! forthcoming0 sheds some light on the accuracy!
relia&ility and e?planatory po*er of representations%
7
I am not suggesting that 1iere and 5an 4raassen ha$e defended the conditions that I descri&e as KsimL
and KisoL &elo*9 &ut they are often understood that *ay% 'his paper can &e seen as articulating clearly
*hat 1iere and 5an 4raassen can &e ta+en to claim and *hat they can(t% I share *ith &oth an emphasis
on the pragmatic dimension of representation: see 1iere -forthcoming0! 5an 4raassen -,22.! ,22I!
,2220%
I
Isomorphism is sometimes said to preser$e! or amount to! <structural identity=% Such terminology is
misleading since the isomorphic structures A and @ are distinct: they ha$e different o&Eects in their
domains% It is rather the <super)structure= of the logical properties of the relations in isomorphic
structures that is identical% 4or that reason I prefer to use the phrase identity of relational frame*or+s%
B
In addition! Duestions of structure)e?emplification are tric+y% ;hich structure is e?emplified &y a
concrete o&Eect is a highly conte?t and purpose dependent issue% Consider for instance! the many
important structures that a &ridge may e?emplify &esides geometric shape: the structure of *eights and
forces! the distri&ution of colours of each of the parts! the relati$e resistances of each part to air and
*ater friction! etc% 'his underdetermination! or plurality! of structure seems to me a maEor o&Eection to
any form of structuralism9 &ut it is someho* tangential to my concerns here%
C
Eric Peterson -unpu&lished manuscript0 argues that the only claim that is essential to the semantic
$ie* is that theories are not linguistic entities% >othing that I ha$e *ritten contradicts that minimal
claim% One might agree that theories are &etter concei$ed as structures! *hile not agreeing that
representation is a structural relation%
8
@lunt -,2C20!*ritten in the midst of the cold *ar! pro&a&ly o$er)emphasises the political aspects of
1uernica% Chipp -,2320! *ritten during the contro$ersy o$er 1uernica(s return to the ne* Spanish
democracy! and in$ol$ed in the international diplomatic efforts that ensued! definitely under)
emphasises them% 'he most &alanced account may remain Arnheim(s -,2C.0%
3
I employ 1uernica to the same effect in SuRreF -,2220% 4rench -forthcoming0 misreports my argument
as one of am&iguity &et*een different targets! and then! confusingly! goes on to *rite in response that
<it is not difficult to find other e?amples from the history of art *hich might &e called non)
representational= -p% B0% Am&iguity is no pro&lem for KisoL! since it is al*ays possi&le for different
o&Eects to e?emplify isomorphic structures% And I neither claim that 1uernica is non)representational9
that *ould &e an a&surd claim for me to ma+e since it *ould &ypass *hat is at sta+e! namely *hether
there can &e representation *ithout isomorphism%
2
'here does not seem to &e a theory of mental intentionality that is free of pro&lems! or has not already
&een refuted% :ere is a sample list: similarity accounts of intentionality are refuted &y! among others!
Cummins -,232! chapter 70 *ho also has strongly criticised co$ariance! or causal accounts -,232!
chapters I)C9 ,2280% Mc"onald -,2230! Egan -,2230 and Milli+an -.///0 offer strong and con$incing
arguments against Cummins( o*n isomorphism theory of representation9 and Milli+an(s -,23I0
adapti$e role theory! *hich is in any case not suited to e?plaining concrete instances of scientific
representation! has &een strongly criticised &y M Cummins -,2280N
,/
'his is not to say that representational forces can not &e studied scientifically% A good deal of historical
and sociological research! for e?ample! is in one *ay or another de$oted to o&Eecti$ely settling issues
of past representational forces! and historians ha$e de$eloped some sophisticated tools to carry out
these tas+s: @a?andall -,23B0! for instance! *as an influential milestone in the history of art% @ut
although science can study $alues! it can not reduce them to facts%
,,
'his is in fact S*oyer(s <penultimate= definition% :is final proposal includes an additional refinement
to account for the further distinction &et*een cases in *hich the representation correlates elements of @
uniDuely to elements of A and those in *hich the representation correlates elements of A uniDuely to
elements of @% Since the distinction is only reDuired to co$er cases of linguistic! or *ord)to)o&Eect
representation! I ignore it here%

Vous aimerez peut-être aussi