On actor-network theory: A few clarifications

Author(s): Bruno Latour

Source: Soziale Welt, 47. Jahrg., H. 4 (1996), pp. 369-381
Published by: Nomos Verlagsgesellschaft mbH
Stable URL: http://www.jstor.org/stable/40878163 .
Accessed: 06/02/2015 04:38

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

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

Nomos Verlagsgesellschaft mbH is collaborating with JSTOR to digitize, preserve and extend access to Soziale
Welt.

http://www.jstor.org

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory
A fewclarifications

Von BrunoLatour

I.

Exploring theproperties ofactor-networks is thetaskthattheParisgroupofscience

andtechnology studieshassetitself totacklesincethebeginning ofthe1980s(Calln/
Law/Rip1986). However, this theory has often been misunderstood andhencemuch
abused.In thispaperI wouldliketo listsomeoftheinteresting properties ofnetworks
andexplainsomeofthemisunderstandings thathavearisen.I willnotconcernmyself
herewiththequantitative studies,especiallytheso-called"co-wordanalysis'/ since
theyarethemselves misunderstood becauseof thedifficulty of exactlygrasping the
socialtheoryand quaintontologyentailedby actor-network (butsee Callon/Cour-
tial/Lavergne 1989a; b).
Threemisunderstandings aredueto commonusagesofthewordnetwork itselfand
theconnotations theyimply.
The firstmistake wouldbe to giveit a commontechnical meaning in thesenseofa
sewage, or train,or subway, or telephone"network". Recenttechnologies oftenhave
thecharacter ofa network, thatis,ofexclusively related
yetverydistant elements with
thecirculation between nodesbeingmadecompulsory through a setofrigorous paths
givingto a fewnodesa strategic character. Nothingis moreintensely connected, more
distant,morecompulsory, andmorestrategically organizedthana computer network.
Suchis nothowever thebasicmetaphor ofan actor-network. A technical network in
theengineer's senseis onlyone ofthepossiblefinalandstabilized statesof an actor-
network. An actor-network maylackall thecharacteristics ofa technical network - it
be
may local, mayit have no compulsory no
paths, strategically positioned nodes. Tom
Hughes's"networks ofpower"(1983),togivea historical example, areactor-networks
atthebeginning ofthestory, andonlysomeoftheirstabilized elements endup being
networks in theengineer's sense,thatis theelectrical grid.Evenat thislaterstagethe
engineering definition ofnetworks is stilla partialprojectionofan actor-network.
The secondmisunderstanding is easytolift:theactor-network theory(henceANT)
hasverylittleto do withthestudyofsocialnetworks. Thesestudies, no matter how
interesting,concernthemselves withthesocialrelations ofindividual humanactors-
theirfrequency, distribution, homogeneity, proximity. It wasdevisedas a reaction to
theoftentoo globalconceptslikethoseof institutions, organizations, statesand na-
tions,addingto thema morerealistic andsmallersetofassociations. Although ANT
sharesthisdistrust forsuchvagueall-encompassing sociologicalterms, it also aimsat
describing theverynatureofsocieties. Butto do so it doesnotlimititselfto human
individualactors,but extendsthe wordactor- or actant- to non-human, non-
individualentities.Whereassocial networkadds information on the relationsof
humansin a socialandnaturalworldwhichis leftuntouched bytheanalysis, ANT
aimsataccounting fortheveryessenceofsocieties andnatures.
Itdoesnotwishto add
socialnetworks to socialtheory, butto rebuildsocialtheoryoutofnetworks. It is as
muchan ontologyor a metaphysics as a sociology(Mol/Law1994).Socialnetworks
SozialeWelt47(1996),S. 369-381

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
370 BrunoLatour

will of coursebe includedin the description,but theywill have no privilegenor pro-

minence(and veryfewof theirquantitativetools have been deemed reusable).
Whythenuse thewordnetwork,sinceit is open to suchmisunderstandings? The use
of the wordcomes fromDiderot. The word "rseau" was used fromthe beginningby
Diderot to describematterand bodies in orderto avoid the Cartesiandivide between
matterand spirit.Thus, the word has had a strongontologicalcomponentfromthe
beginning(Anderson1990).Put too simply,ANT is a changeof metaphorsto describe
essences:insteadof surfacesone gets filaments(or rhyzomesin Deleuze's parlance
Deleuze/Guattari1980)). More preciselyit is a changeof topology.Insteadof thinking
in termsofsurfaces- two dimensions- or spheres- -threedimensions- one is asked
to thinkin termsof nodes thathaveas manydimensionsas theyhave connections.As
a firstapproximation,ANT claimsthatmodernsocietiescannotbe describedwithout
recognizingthem as having a fibrous,thread-like,wiry, stringy,ropy, capillary
characterthat is nevercapturedby the notions of levels,layers,territories, spheres,
categories,structures, systems.It aims at explainingthe effects
accountedforby those
traditionalwordswithouthavingto buy the ontology,topology and politicsthatgo
with them.ANT has been developed by studentsof science and technology,and its
claim is thatit is utterlyimpossibleto understandwhat holds societytogetherwithout
reinjectingin its fabricthe factsmanufacturedby naturaland social sciencesand the
artefactsdesignedby engineers.As a second approximation,ANT is thus the claim
thattheonly wayto achievethisreinjectionof thingsintoour understanding of social
fabricsis througha network-like ontology and social theory.
To remainat thisveryintuitivelevel,ANT is a simplematerialresistanceargument.
Strength does not come fromconcentration, purityand unity,butfromdissemination,
heterogeneity and the carefulplaitingof weak ties. This feelingthat resistance,ob-
duracy and sturdinessare more easily achieved throughnetting,lacing, weaving,
twistingof tiesthatare weak by themselves,and thateach tie,no matterhow strong,
is itselfwovenout of stillweakerthreads,permeatesforinstanceFoucault's analysisof
micro-powersas well as recent sociology of technology.But the less intuitive
philosophicalbasisforacceptingANT is a background/foreground reversal:insteadof
startingfromuniversallaws - social or natural- and takinglocal contingenciesas so
manyqueer particularities thatshould be eithereliminatedor protected,it startsfrom
irreducible,incommensurable,unconnectedlocalitieswhich then, at a greatprice,
sometimes end into provisionally commensurable connections. Through this
foreground/background reversalANT has some affinity withtheorderout ofdisorder
or chaos philosophy(Serres1983; Prigogine/Stengers 1979) and manypracticallinks
withethnomethodology (Garfinkeland Lynch'sprinciplein Lynch 1985). Universali-
ty or orderare not the rule butthe exceptionsthathaveto be accountedfor.Loci, con-
tingencies or clustersare more like archipelagoson a sea thanlike lakesdottinga solid
land. Less metaphorically, whereasuniversalists havetofill in thewhole surfaceeither
with orderor with contingencies, ANT does not attemptto fillin what is in between
local pocketsof ordersor in betweenthe filamentsrelatingthesecontingencies.

This is the mostcounter-intuitive aspectof ANT. Literallythereis nothingbut net-

works, there is nothing in between them,or, to use a metaphorfromthe historyof
physics,there is no aetherin which networks should be immersed.In thissenseANT
is a reductionistand relativisttheory,but, as I shall demonstrate,this is the first
necessarysteptowardsan irreductionist and relationistontology.

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory 371

IL

ANT makesuse of some of the simplestpropertiesof netsand thenadds to it an actor

thatdoes some work;the additionof such an ontologicalingredientdeeplymodifiesit.
I will startout with the simplestpropertiescommon to all networks.
Far/close:the firstadvantageof thinkingin termsof networksis thatwe get rid of
*
"the tyrannyof distance*or proximity.Elementswhich are close when disconnected
may be infinitelyremotewhen theirconnectionsare analyzed; conversely,elements
which would appear as infinitelydistantmay be close when their connectionsare
broughtback into the picture.I can be one metreaway fromsomeone in the next
telephonebooth and nevertheless be morecloselyconnectedto mymother6000 miles
away;an Alaskan reindeermightbe tenmetresawayfromanotherone and theymight
nevertheless be cut offby a pipeline of 800 milesthatmakestheirmatingforeverim-
possible;myson maysitat school witha youngarabof his age,but in spiteofthisclose
proximityin firstgradetheymightdriftapartin worldsthatwill become incommen-
surablelater;a gas pipe maylie in the groundclose to a cable televisionglassfiberand
nearbya sewage pipe, and each of them will neverthelesscontinuouslyignore the
parallelworldslyingaroundthem.The difficulty we have in definingall associations
in termsof networksis due to the prevalenceof geography.It seems obvious thatwe
can oppose proximityand connections.However,geographicalproximityis the result
of a science- geography-, of a profession- geographers -, of a practice- mapping
system,measuring,triangulating. Their definitionof proximityand distanceis useless
forANT - or it should be includedas one typeof connections,one typeof networks,
as we will see below.All definitionsin termsof surfaceand territories come fromour
readingof mapsdrawnand filledin by geographers. Out of geographers and geography,
"in between"theirown networks,thereis no such thingas a proximityor a distance
which would not be definedby connectibility.The geographicalnotion is simply
anotherconnectionto a griddefininga metricsand a scale (Jacob 1990). The notion
of networkhelpsus to liftthe tyrannyof geographersin definingspace and offersus
a notion which is neithersocial nor "real" space, but associations.
Small scale/largescale: the notion of networkallows us to dissolvethe micro-macro-
distinctionthathas plagued social theoryfromitsinception.The whole metaphorof
scales going fromthe individualto the nation state,throughfamily,extendedkin,
groups,institutionsetc.is replacedby a metaphorof connections.A networkis never
biggerthan anotherone, it is simply longeror moreintenselyconnected.The small
scale/largescale model has three featureswhich have proven devastatingfor social
theory:it is tiedto an orderrelationthatgoes fromtop to bottomor frombottomto
top - as ifsocietyreallyhad a top and a bottom;it impliesthatan element"b" being
macro-scaleis of a different natureand should thusbe studieddifferently froman ele-
ment"a" whichis micro-scale;it is utterlyunable to followhow an elementgoesfrom
being individual- a - to collective- b - and back.
The networknotion impliesa deeplydifferent social theory:it has no a prioriorder
relation;it is not tiedto theaxiologicalmythof a top and a bottomof society;it makes
absolutelyno assumptionwhethera specificlocus is macro-or micro-and does not
modifythe tools to studyelement"a" or element"b"; thus,it has no difficulty in
followingthe transformation of a poorly connectedelementinto a highlyconnected
one and back.The networknotion is ideallysuitedto followthechangeof scales,since
it does not requirethe analystto partitionherworldwithany a prioriscale.The scale,
that is, the type, number and topographyof connections, is left to the actors
themselves.

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
372 BrunoLatour

The notion of networkallowsus to liftthe tyrannyof social theorists,to regainsome

marginof manoeuvresbetweentheingredients ofsociety- itsverticalspace,itshierar-
-
chy,its layering,its macro-scale,its wholeness,its overarchingcharacter and to see
how thesefeaturesare achievedand what stufftheyare made of. Insteadof havingto
choose betweenthelocal and theglobalview,thenotionof networkallows us to think
of a globalentity- a highlyconnectedone - whichnevertheless remainscontinuous-
ly local . . . Insteadof opposingthe individuallevelto the mass,or agencyto structure,
we simplyfollowhow a givenelementbecomesstrategicthroughthe numberof con-
nectionsit commands,and how it loses its importancewhen losingits connections.

Inside/outside: thenotionofnetworkallowsus to getridofa thirdspatialdimension

afterthoseof far/closeand big/small.A surfacehas an insideand an outsideseparated
by a boundary.A networkis all boundarywithoutinsideand outside.The only ques-
tion one mayask is whetheror not a connectionis establishedbetweentwo elements.
The surface"in between"networksis eitherconnected- but thenthe networkis ex-
panding- or non-existing. Literally,a networkhas no outside.It is not a foreground
overa background,nor a crackonto a solid soil, it is like Deleuze's lightningrod that
createsby the same strokethe backgroundand the foreground(Deleuze 1968). The
greateconomyof thinkingallowedby the notion of networkis thatwe are no longer
obligedto fillin the space in betweenthe connections- to use a computermetaphor,
we do not needthelittlepaintbox familiarto MacPaintusersto "fillin" theinterspace.
A networkis a positivenotionwhich does not need negativity to be understood.It has
no shadow.

The notionofnetwork,in itsbaresttopologicaloutline,alreadyallowsus to reshuffle

spatialmetaphorsthathaverenderedthe studyof society-nature close and
so difficult:
far,up and down, local and global,insideand outside. They arereplacedby associations
and connections(whichANT does not haveto qualifyas beingeithersocial or natural
or technicalas I will show below). This is not to saythatthereis nothinglike "macro"
societyor "outside" natureas ANT is oftenaccusedof,but thatin orderto obtainthe
effectsof distance,proximity, hierarchies,connectedness,outsidernessand surfacesan
enormoussupplementary work has to be done (Latour 1996a). This work,however,is
not capturedby the topologicalnotion of network,no matterhow sophisticatedwe
wish to makeit. This is whyANT adds to the mathematicalnotion of networka com-
pletelyforeignnotion,thatof actor.The new hybrid"actor-network"leads us away
frommathematicalpropertiesinto a world which has not yetbeen so neatlycharted.
To sketchthesepropertieswe should now moveon fromstaticand topologicalproper-
ties to dynamicand ontologicalones.

III.

A networkin mathematicsor in engineeringis somethingthatis tracedor inscribed

by some otherentity- the mathematician, the engineer.An actor-network is an enti-
ty thatdoesthetracing and the It
inscribing. is an ontologicaldefinition
and not a piece
of inertmatterin thehandsof others,especiallyofhumanplannersor designers.It was
in orderto point out thisessentialfeaturethatthe word "actor" was added to it.

Now, the word actorhas been open to the same misunderstanding as the word net-
work. "Actor" in the Anglo-Saxontraditionis alwaysa human intentionalindividual
actorand is mostoftencontrastedwith mere"behaviour".If one adds thisdefinition
of an actorto thesocial definitionof a network,thenthe bottomof misunderstanding
is reached:an individualhuman- usuallymale - who wishesto grabpowermakesa

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory 373

networkof allies and extendshis power- doing some "networking"or "liaising" as

the Americanssay ... Alas, thisis the way ANT is most oftenrepresented, which is
about as accurateas sayingthatthe nightskyis black because the astrophysicists have
shownthereis a big black hole in it. An "actor" in ANT is a semioticdefinition- an
actant-, thatis somethingthatactsor to whichactivityis grantedby others.It implies
no specialmotivationofhumanindividualactors,nor of humansin general.An actant
can literallybe anythingprovidedit is grantedto be the sourceof an action.Although
thispointhas been made overand overagain,anthropocentrism and sociocentrismare
so strongin social sciences(as well as in the critiquesof social explanations)thateach
use of ANT has been construedas ifit talkedof a fewsuperhumanslongingforpower
and stoppingat nothingto achieve theirruthlessgoals . . . Even my own network
studyofPasteur(Latour 1988a) - in spiteofthelengthyontologicalsecondpart- has
oftenbeen understoodas a Madison Avenueversionof science- which is unfairnot
only to my account but also to Madison Avenue ... If a criticismcan be levelledat
ANT it is, on the contrary, its completeindifference forprovidinga model of human
competence. There is no model of (human) actor in ANT nor any basic list of com-
petencesthathaveto be setat the beginning,becausethe human,the selfand thesocial
actorof traditionalsocial theoryis not on its agenda.
So whatis on itsagenda?The attributionofhuman,unhuman,nonhuman,inhuman
characteristics;the distributionof propertiesamong these entities;the connections
establishedbetweenthem;the circulationentailedby theseattributions, distributions
and connections;the transformation of those attributions,distributionsand connec-
tionsof the manyelementsthatcirculate,and of the fewwaysthroughwhich theyare
sent.
The difficultyof graspingANT is thatit has been made bythefusionofthreehither-
to unrelatedstrandsof preoccupations:
- a semioticdefinitionof
entitybuilding;
- a
methodologicalframeworkto recordthe heterogeneity of such a building;
- an ontologicalclaim on the
"networky" characterof actantsthemselves.
ANT assertsthatthelimitsofthesethreeunrelatedinterests aresolvedwhen,and on-
ly when, theyare fusedinto an integrated practiceof study.
Semioticsis a necessarystepin thisventure,sincewhen you bracketout thequestion
of referenceand that of the social conditionsof productions- that is nature"out
there" and society"up there"- what remainsis, in a firstapproximation,meaning
production,or discourse,or text.This is the major achievementof the sixtiesand of
their"linguisticturn"or "semioticturn" . Insteadof beingmeansof communications
betweenhuman actorsand nature,meaningproductionsbecame the only important
thingto study.Insteadof beingunproblematic,theybecame opaque. The taskwas no
longerto make them more transparent, but to recognizeand relishtheirthick,rich,
layeredand complex matter.Instead of being mere intermediaries, theyhad become
mediators.From a means,meaninghas been made an end in itself.For twentyyearsthe
bestmindshavebeen busyexploringall theconsequencesofthismajorturnawayfrom
thenavemodel of communication.Their oftenstructuralist interpretationshavebeen
dismantled,but what remainsis a toolbox to studymeaningproductions.ANT sorts
out fromthis toolbox what is usefulto understandthe constructionof entities.The
keypoint is thateveryentity,includingthe self,society,nature,everyrelation,every
action,can be understoodas a "choice" or a "selection" of finerand finerembranch-
ments going from abstractstructure- actants- to concreteones - actors. The
generativepaththatis thusretracedgivesan extraordinary libertyof analysiscompared

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
374 BrunoLatour

to theempoverished"social vocabulary"thatwas used earlier- and is now in fashion

again. Of course the structuralrenderingof these choices - differences- and
embranchments- dichotomies- are not kept by ANT, but essential traitsof
semioticsare kept.First,thegrantingof humanityto an individualactor,or the gran-
ting of collectivity,or the grantingof anonymity,of a zoomorphic appearance,of
amorphousness,of materiality, requirespayingthesamesemioticprice.The effects will
be different, the genreswill be different, but not the workof attributing, imputing,
distributing action,competences,performances and relations.Secondly,actorsare not
conceived as fixedentitiesbut as flows,as circulatingobjects undergoingtrials,and
theirstability,continuity,isotopyhas to be obtainedby otheractionsand othertrials.
Finally,what is keptfromsemioticsis the crucialpracticeto granttextsand discourses
the abilityto definealso theircontext,theirauthors- in the text-, theirreaders-
in fabula- and eventheirown demarcationand metalanguage.All theproblemsofthe
analystareshiftedto the"textitself"withouteverbeingallowedto escapeintothecon-
text(Greimas 1976). Down with interpretation! Down with the context!The slogans
of thethe60s and 70s "everything is a text","thereis only discourse","narrativesexist
bythemselves", "we haveno accessto anythingbutaccounts" arekeptin ANT butsav-
ed fromtheirontologicalconsequences.This salvation,however,does not come byfall-
ingback on thepre-deconstruction common sense- "afterall, thereis a social context
up thereand a natureout there"- but by extendingthe semioticturnto thisfamous
natureand thisfamouscontextit has bracketedout in the firstplace.

A major transformation of these slogans occurredwhen semioticswas turnedto

scientificand technicaldiscourseby ANT - and especiallyto scientifictexts.As long
as one studiedfictions,myths,popular cultures,fashions,religions,politicaldiscourse,
one could hold to the "semioticturn" and takethemas so many"texts".Scholarsdid
not seriouslybelievein themanyway,and thusthe intellectualdistanceand scepticism
was easyto achievewhile thedouble treasuryof "scientism"and "socialism" was kept
intactin theirheart.But what about scientifictruthand materialefficiency? What
about the reference"out there" in hard scientifictexts?This was the real testfor
semiotics,and althoughit passed the triala price had to be payed.In the practiceof
ANT semioticswas extendedto definea completelyempty framethat enabled to
followany assemblageof heterogeneousentities- includingnow the "natural" en-
titiesof scienceand the "material"entitiesof technology.This is the second strandof
ANT: it is a methodto describethe deploymentof associationslike semiotics;it is a
methodto describethegenerative pathof anynarration.It does not sayanythingabout
the shape of entitiesand actions,but only what the recordingdevice should be that
would allow entitiesto be describedin all theirdetails.ANT places the burdenof
theoryon the recording, not on the specificshape thatis recorded.When it saysthat
actorsmaybe humanor unhuman,thattheyareinfinitely pliable,heterogeneous, that
theyare freeassociationists,know no differences of scale,thatthereis no inertia,no
order,thattheybuild theirown temporality, thisdoes not qualifyanyrealobservedac-
tor,but is thenecessaryconditionfortheobservationand therecordingof actorsto be
possible. Instead of constantlypredictinghow an actor should behave and which
associationsare alloweda priori,ANT makesno assumptionat all, and in orderto re-
main uncommittedit needsto set its instrument by insistingon infinitepliabilityand
absolutefreedom.In itselfANT is not a theoryof action,no more than cartography
is a theoryon the shape of coast lines and deep sea ridges;it just qualifieswhat the
observershould suppose in orderforthe coast linesto be recordedin theirfinefractal
patterns.Any shape is possible provided it is obsessivelycoded as longitude and

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory 375

latitude.Similarly,any association is possible provided it is obsessivelycoded as a

heterogeneousassociation throughtranslations.It is more an infralanguagethan a
metalanguage.It is evenless thana descriptivevocabulary;it simplyopens,againstall
a priorireductions,the possibilityof describingirreductions(Latour 1988a, partII).
ANT is not merelyempiricistthough,since in order to definesuch an irreducible
space in whichto deployentitiessturdytheoreticalcommitmentshave to be made and
a strongpolemical stancehas to be taken,so as to forbidthe analystto dictateactors
whattheyshoulddo. Such a distributionofa strongtheoryfortherecordingframeand
no middle range theory for the the description is another source of many
misunderstandings, since ANT is accused of eitherbeing dogmaticor of providing
meredescriptions.For the same reason it is also accused of claimingthat actorsare
"really" infinitelypliable and freeor,inversely, of not tellingwhat a human actorreal-
ly is after(Lee/Brown1994).
The firsttwo strands- the semioticand the methodologicalone - by themselves
will be open to criticism.The firstbecausethereis no way to considerthatbracketing
out social contextand reference solvesthe problemof meaning- in spiteof the now
dated claims of the swingingseventies-, and the second because merelydeploying
shapesof associationsmightbe a worthwhiledescriptivetaskbutdoes not offeranyex-
planation.It is only when a thirdstrandis added to those two and ontologicalclaims
on networksare made thatANT escapescriticism.This move,however,is so devious
thatit has escapedthe attentionof manyusersof ANT. Which is a pity,since once it
is made,ANT loses its radicalcharacterand soon appearscommensensicalenough.
The weaknessof semioticshas always been to considermeaningproductionaway
fromwhat the natureof entitiesreallyis; when semioticsis turnedto naturehowever
and unhumanentitiesare allowed to enterinto the picture,it soon appears thatthe
words"discourse"or "meaning" maybe droppedaltogetherwithoutanydangerofgo-
ing back to nave realismor nave naturalism.It is only because semioticiansstudied
texts- and literaryones at that - insteadof thingsthat theyfeltobliged to limit
themselvesto "meaning". In effectthey scientisticallybelieved in the existenceof
thingsin additionto meaning(not mentioningtheirbeliefin the existenceof a good
old social contextwheneverit suitedthem).But a semioticsof thingsis easy,one simp-
ly has to drop the meaningbit fromsemiotics. . .
Ifone now translates semioticsbypath-building or order-making or creationofdirec-
tions,one does not have to specifyifit is languageor objectsone is analyzing.Such a
move givesa new continuityto practicesthatwere deemed different when one dealt
withlanguageand "symbols",or with skills,work and matter.This move can be said
eitherto elevatethingsto thedignityof textsor to elevatetextsto the ontologicalstatus
ofthings.What reallymattersis thatit is an elevationand not a reduction,and thatthe
new hybridstatusgivesto all entitiesboth the action,varietyand circulatingexistence
recognizedin the studyof textualcharactersand the reality,solidity,externality that
was recognizedin things"out of" our representations. What is lost is the absolute
distinctionbetweenrepresentation and things- but thisis exactlywhat ANT wishes
to redistributethrough what I have called a counter-copernican revolution.
Once settledthisfirstsolution- extendingsemioticsto thingsinsteadof limitingit
to meaning-, the second difficulty fallswith it - buildingan emptymethodological
frameto register description.Actor-networks do connect,and by connectingwithone
anotherprovidean explanationof themselves,the only one thereis forANT. What is
an explanation?The attachmentof a set of practicesthatcontrolor interfere in one
another.No explanationis strongeror more powerfulthan providingconnections
amongunrelatedelementsor showinghow one elementholds manyothers.This is not

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
376 BrunoLatour

a propertythatis distinctfromnetworksbut one of theiressentialproperties(Latour

1988b). They become moreor less explanableas theygo and dependingon what they
do to one another.Actorsare cleaningup theirown mess,so to speak. Once you grant
themeverything, theyalso giveyou back the explanatorypowersyou abandoned.The
very divide between descriptionand explanation,hows and whys,blind empiricism
and hightheorizingis as meaninglessforANT as the difference betweengravitation
and space in relativitytheory.Each network,by growing,"binds" the explanatory
resourcesaround it, and thereis no way theycan be detachedfromits growth.One
does not jump outsidea networkto add an explanation- a cause,a factor,a set of fac-
tors,a seriesof co-occurences;one simplyextendsthe networkfurther. Everynetwork
surroundsitselfwithitsown frameof reference, itsown definitionof growth,of refer-
ing,of framing, of explaining.In thisprocesstheframeof reference of the analystdoes
not disappearmorethan the physicist'sin Einstein'sworld; on the contrary, at last it
is able to extenditself,but at a price: the framebecomes,as it does in GeneralRelativi-
ty,"a molluscof reference"insteadof a detachedGalilean frame,and each accounthas
to be recalculatedby the ANT equivalentof a Lorentztransformation (Latour 1988c).
There is no way to providean explanationifthe networkdoes not extenditself.This
is not in contradictionwiththe scientifictaskof providingexplanationand causality,
sincewe learnedfromtheverystudiesofhardsciencesthatno explanationofanyscien-
tificphenomenonand no causalitycould be providedwithoutextendingthe network
itself.By tyingthe explanationto the networkitselfANT does not abandon the goal
of science,sinceit showsthatthisgoal has neverbeen achieved,at leastnot throughthe
epistemologicalmythof explanation.ANT can't depriveitselfof a resourceit shows
no one has everhad in the firstplace. Explanationis ex-plicated, thatis unfolded,like
gravityin Einstein's curved space, it is still there as an effect,but it is now
indistinguishable fromthe description,the deploymentof the net.

position- but one should preferthe less loaded termof relationist

This relativistic
- solvestwo otherproblems:that of historicityand thatof reflexivity.

The pre-relativist debatebetweenprovidingan explanationand "simply" documen-

tingthe historicalcircumstances fallsapart:thereis no differencebetweenexplaining
and tellinghow a networksurroundsitselfwith new resources;if it "escapes socio-
historicalcontingencies",as criticsoftenargue,then this simply means that other,
longerlastingresourceshave been garneredto stay around - the etymologyof cir-
cumstances.Hughes's networksof power grow (Hughes 1983), and by their very
growththeybecome moreand moreof an explanationof themselves;you do not need
an explanationfloatingoverthemin additionto theirhistoricalgrowth;Braudel'snet-
worksand worldeconomicsgrow,and theyarewhatthe "big causes" are made of.You
do not need to add to themCapitalismor Zeitgeist exceptas anothersummary,another
punctualisationofthenetworksthemselves.Eitherthecause designatesa body ofprac-
ticeswhich is tiedto the networkunderdescription- and thisis what growthof net-
work means-, or it is not related,and thenit is just a word added to the description,
literallyit is the word"cause". In thissense,ANT giveshistoryits legitimateplace -
which is not theplace prudenthistorianslike to sit on, as safelyaway as possiblefrom
ontologicalquestions.There is nothingbetter,sturdierthan a circumstancialdescrip-
tion of networks."It just happensto be thisway".
But such a summarywould be construedas historicismifit werenot understoodas
a definitionof the thingsthemselves.The debatebetweenhistoricismand explanation
or theorywas not solvableas long as therewas, on the one hand, a historyof people,
of contingencies,of what is "in time" and, on the otherhand,a theoryor a scienceof

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory 377

what is timeless,eternal,necessary.For ANT thereis scienceonly of the contingent,

as of necessityit is locally achievedonly throughthe growthof a network.If thereis
also a historyof things,then the debate between descriptionand ex-planation,or
historicityand theory,is entirelydissolved.For ANT this is not the proof of the
weaknessof its explanatorypowers,since describingor accountingfora networkis
what an explanationor an explicationis and what has alwaysbeen the case in the so-
called hard sciences - or more exactly "progressivelyhardenedsciences" (Latour
1996b).
Althoughnot the main goal of ANT, reflexivity is added as a bonus once the frames
of referenceare grantedback to the actors- and once the actorsare grantedback the
possibilityof crossingthe sacred dividingline between thingsand representations
(Ashmore1989).Reflexivity is seen as a problemin relativisttheory,becauseit appears
thateitherthe observerrequestsa statusit deniesto others,or it is as silentas all the
othersto which any privilegedstatusis denied. This "problem" falls,however,when
the epistemolgica!mythof an outsideobserverprovidingan explanationin addition
to "mere description"disappears.There is no longer any privilege- but therehas
neverbeen any need forit either.The observer- whateverit is - findsitselfon a par
withall the otherframesof reference. It is not leftto despairor navel-gazing,sincethe
absenceof privilegedstatushas neverlimitedthe expansionand intelligenceof any ac-
tor.Worldbuilderamongworldbuilders,itdoes not see a dramaticlimiton knowledge
in itsabandonmentof Galilean frames,but only resources.To extendfromone frame
of reference to the nextit has to work and pay the price like any otheractor.In order
to explain,to account,to observe,to prove,to argue,to dominateand to see it has to
move around and work (I should say it has to "network" ). No privilegealso means
no a priorilimitson knowledge.If actorsare able to accountforothers,so can it. If ac-
torscan't,it mightstilltry.History,risksand venturesare also in the observers'own
networkbuilding.Such is ANT's solution to reflexivity (Stengers1993).
Reflexivityis not a a
"problem", stumbling block along the path to knowledge,the
prison in which all enterpriseswould be locked, it is the land of opportunityat last
opened to actorswhich are primusinterpares,or striveforparityor primacylike any
other.Of course reusablemetalangageis abandoned,but this is not givingup much,
sinceobserverswho displayedtheirrichmetalanguagewereusallysmallpointslimited
to veryspecificloci - campuses,studios,corporaterooms. The price ANT pays to
move fromone locus to the nextis thatthereare as manymetalanguagesas thereare
framesof reference - the only
metalanguagerequired(see above strand2) beingmore
adequatelycalled an infralanguage which has to be poor, limited,shortand simple-
the equivalentof a Lorentztransformation beingcalled "translation"in ANT (Latour
1988c). This infralanguageis enough to move from one net to the other,and the
specificexplicationwill alwaysbe a one-shotaccount exclusivelytailoredto the pro-
blem at hand (Lynch's principle,Callon's "disposable explanations",Serres's"cross
over betweenexplanandumand explanans" (Serres1995)). If it is more generallyap-
plicable,it meansthatit is ridingover a networkthatexpandsitself.
This solution becomes commonsenseonce it is acceptedthat an account or an ex-
plicationor a proof is alwaysadded to the world; that it does not subtractanything
fromtheworld.Reflexivists as well as theirpre-relativist enemiesdreamof subtracting
knowledge from the in
things themselves.ANT keepsaddingthingsto the world,and
its selectionprincipleis no longerwhetheror not thereis a fitbetweenaccount and
reality- thisdual illusionhas been dissolvedaway -, but whetheror not one travels
froma net to another.No metalanguageallows you to do thistravel.By
abandoning
the dreamsof epistemologyANT is not reducedto moralrelativism,but getsback a

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
378 BrunoLatour

stronger deontologicalcommitment: eitheran accountleadsyouto all theotherac-

counts- anditis good-, or itconstantly themovement,
interrupts frames
letting of
reference distant
andforeign - anditis bad.Eitheritmultipliesthemediating points
between anytwoelements - anditis good-, or itdeletesandconflates mediators -
anditis bad.Eitheritis reductionist- andthat'sbadnews-, or irreductionist - and
that'sthehighest ethicalstandard forANT. Wewillseethatthistouchstone is much
morediscriminating thanthequestforepistemological purity,or forfoundations,or
formoralnorms.Demarcation is in factan enemyofdifferentiation.
Building on thesemioticturn,ANT first bracketsoutsocietyandnaturetoconsider
onlymeaning-productions; then,breaking withthelimitsofsemiotics withoutlosing
itstoolbox,itgrants activitytothesemioticactorsturning themintonewontological
hybrids, worldmakingentities;by doingsucha counter-copernican revolutionit
buildsa completely emptyframe fordescribing howanyentity buildsitsworld;final-
ly,it retainsfromthedescriptive projectonlyveryfewterms- itsinfralanguage -
whicharejustenoughtosailin between framesofreference,
andgrants backto theac-
torsthemselves theabilityto buildpreciseaccountsofone anotherbytheveryway
theybehave;thegoalof buildingan overarching explanation- thatis,forANT, a
centre ofcalculation
thatwouldholdorreplaceorpunctuate alltheothers- isdisplac-
ed bythesearchforex-plications, thatis forthedeployment ofas manyelements as
possibleaccounted forthrough as manymetalanguages as possible.

IV.

Now thatthebasictopologicalproperties ofnetworks havebeensketched- se-

condsection- and thatthebasicontologicalfeatures ofactorshavebeenoutlined
- sectionabove-, thereis no difficulty in seeingthatANT is notabouttraced net-
works,butabouta network-tracing activity.As I saidabove,thereis nota netandan
actorlayingdownthenet,butthereis an actorwhosedefinition oftheworldoutlines,
traces,
delineates, describes, files,lists,records, marksor tagsa trajectory thatis called
a network. No netexistsindependently oftheveryactoftracing it,andno tracing is
donebyan actorexterior tothenet.A network is nota thing, buttherecorded move-
mentof a thing.The questionsANT addresses now havechanged.It is no longer
whether a netis a representation or a thing,a partofsocietyor a partofdiscourse or
a partofnature, butwhatmovesandhowthismovement is recorded.
Wecannotsaythatwhatmovesinsidenetworks arepiecesofinformation, genes,cars,
bytes,salutations, words,forces, opinions,claims,bodies,energy etc.,sinceANT also
wantstoreconstruct netsbefore thereis anydistinction between whatcirculates inside
andwhatkeepsthemon track,so to speak,fromtheoutside.Again,as I saidat the
beginning, thetechnical metaphor ofnetworks is a latecomer forANT anddoesnot
capturethetracing activity. No, whatcirculates hasto be defined likethecirculating
objectin the semiotics of texts- especially scientifictexts(Bastide1990).It is defined
bythecompetence it is endowedwith,thetrialsit undergoes, theperformances it is
allowedtodisplay, theassociations itis madetobearupon,thesanctions itreceives,the
background inwhichitiscirculating etc.Itsisotopy - thatisitspersistence intimeand
space- is nota property ofitsessence, buttheresultofthedecisionstakenthrough
thenarrative programmes andthenarrative paths.
However,such a classicdefinition would limitANT to the worldof textand
discourse.Whathappenswhena circulating objectleavestheboundary ofa text?The
traditionalansweris thatthereis a yawning gap in between the text and thecontext.
At theinterface a dramatic trialis supposedto abruptly intervene through whichthe

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory 379

circulatingobject is assessedeitherby checkingits referential

fitor its social interest.
Not so forANT, whichdoes not believein thisdistinction,sinceit has extendedmean-
ing productionsto all productions.For ANT the gap is no more than a slightbump
alongthenet;theyawnis an artefactcausedby a previousdividebetweennature,socie-
ty and discourse.For ANT on the contrary,thereis a continuity,a multiplicityof
plugs betweenthe circulatingobjects in the text,the claims outside the textin the
"
"social", and what the actantsthemselvesreallydo in nature* ' The circulatingobject
goes on circulatingand goes on gettingits isotopyfromwhat other actorsdo to it.
"Society" has the same net-likepropertiesas have texts,and so has "nature".But it
would be moreaccurateforANT to saythatthesethreecategoriesare arbitrary cutting
points on a continuoustracingof action, and still more accurateto show how these
categoriesthemselvesarepartof the manytrialsand eventsand resourcesthatare used
along the pathsto attribute"textuality"or "sociality" or "naturality"to thisor that
actor.They are partof what is distributed- not partof what makesthe distribution.
There is no off-the-shelf word to describethiscommon movement.To say thatit is
a generalizednarrativepath would immediatelymean that texts are extendedto
everything; to saythatit is a forceor an energyor a geneor a culture-gene would mean
thateverything would be naturalized,includingsocietyand discourse;to say thatit is
a social interest,a social action or labour would extendsocietyto natureand to texts.
It was to get out of thisessentialdifficulty thatANT playedwith a generalizedsym-
metry(Calln 1986) and made a principleof usingwhicheverwordsare connotedin
one of the formerrealmsto describethe others,thus showingthe continuityof net-
worksand the completedisregardforthe artefactualgaps introducedby pre-relativist
arguments.However, this solution is rathertricky,since it may combine all the
misunderstandings - and thisis indeedwhathappenedto ANT, readersand usersalike
saying at once that it is a social constructivist
argument,the returnof naturalism,or a
typically French belief in the overallextension of texts. . . Which of course it is in a
sense,butonly insofaras ANT is thesimultaneousrejectionof naturalisation, socialisa-
tionand textuzlisation. ANT claimsthatthese"(x)-isations"haveto be dissolvedall at
once and thatthejob is not done betterifone of themgainshegemonyor ifthe three
are carefullycircumscribed.All (x)-isationsare the fillingin of what is "in between"
the networks;and which one is chosen or rejectedmakesno practicaldifference, since
netshave no "in between" to be filledin.
Ifchoosingwordsforthenetwork-tracing activityhas to be done,quasi-objects
(Serres
1987) or tokensmightbe the bestcandidatesso far.It is crucialforthe definitionof the
termthatwhat circulatesand what makesthe circulationbe both co-determinedand
transformed. A ball goingfromhand to hand is a poor exampleof a quasi-object,since,
althoughit does tracethe collectiveand althoughthe playingteam would not exist
withoutthemovingtoken,thelatteris not modifiedby thepassings.Conversely,what
I calledthefirstprincipleof sciencestudies(Latour 1987) - thata claim is in thehands
of others- is equally an approximation,sinceit entailshumanlocutorsendowedwith
handsand mouthswho pass a claim withoutthemselvesundergoingdramaticchanges.
As a rule,a quasi-objectshould be thoughtof as a movingactantthattransforms those
who do the moving,because theytransform the movingobject. When the token re-
mains stable or when the moversare kept intact,these are exceptionalcircumstances
which have to be accountedfor.This definitionof what is the rule and what are the
exceptionswould be enoughto tellANT fromall models of communicationsthat,on
thecontrary, beginwithwell definedmoversand movingobjectsand view obstaclesto
exchanges as so manyexceptionsto be explained.But anotherfeatureforbidsany con-
fusionofANT withhuman-centered, or language-centered, or praxis-centeredmodels.

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
380 BrunoLatour

As a rule,what is doingthe movingand what is movedhave no specifichomogeneous

morphism.They can be anthropo-morphic, but also zoo-morphic,physi-morphic,
logo-morphic,techno-morphic, ideo-morphic,thatis "(xj-morphic".It mighthappen
thata generative path has limitedactantsto a homogeneousrepertoire of humansor of
mechanismor of signsor of ideas or of collectivesocial entities,but theseare excep-
tions which should be accountedfor(Latour 1996c).
ANT is a powerfultool to destroyspheresand domains, to regain the sense of
heterogeneity, and to bringinterobjectivity back into the centreof attention(Latour
1994).Yetitis an extremelybad tool fordifferentiating associations.It givesa black and
white picture,not a coloured and contrastedone. Thus it is necessary,afterhaving
tracedthe actor-networks, to specifythe types of trajectoriesthat are obtained by
highly differentmediations. task,and the one thatwill make ANT
This is a different
scholarsbusy fora numberof yearsto come.

Literaturverzeichnis

Anderson,Wilda (1990): Diderot's Dream, The JohnsHopkin's UniversityPress,Baltimore.

Ashmore, Malcolm (1989): The Reflexive Thesis. Wrighting(sic) Sociology of Scientific
Knowledge,Chicago UniversityPress,Chicago.
Bastide,Franoise (1990): The Iconographyof ScientificTexts: Principle of Analysis,in: M.
Lynch, S. Woolgar (Hrsg.), Representationin ScientificPractice,MIT Press, Cambridge
Mass., S. 187-230.
Calln, Michel (1986): Some elements of a sociology of translation:domesticationof the
scallops and the fishermenof St Brieux Bay,in: J. Law (Hrsg.), Power,Action and Belief.A
New Sociology of Knowledge?,Routledee and Regan Paul, London, S. 196-229.
Calln, Michel, Law, John,Rip, Arie (Hrsg.) (1986): Mapping the Dynamics of Science and
Technology,Macmillan, London.
(1989): Co-WordAnalysis: A Tool forthe Evaluation of
Calln, Michel, Courtial,Jean-Pierre
Public, Report forthe NSF grantPRA N85 12 982, Paris.
Lavergne,Francise (1989): La Mthode des motsassocis.
Calln, Michel, Courtial,Jean-Pierre,
Un outil pour l'valuation des programmespublics de recherche.Etude pour la National
Science Foundation,Ecole des Mines, Paris.
Deleuze, Gilles (1968): Diffrenceet rptition,PUF, Paris.
Deleuze, Gilles, Guattari,Flix (1980): Mille plateaux. Capitalisme et schizophrnie,Minuit,
Paris.
Garfinkel,Harry (1967): Studiesin Ethnomethodology,PrenticeHall, New Jersey.
Greimas,AlgirdasJulien(1976): On Meaning. SelectedWritingsin SemioticTheory,University
of Minnesota Press,Minneapolis.
Hughes,Thomas P. (1983): Networksof Power.ElectricSupply SystemsIn the US, Englandand
Germany,1880-1930, The JohnHopkins UniversityPress,Baltimore.
Jacob,Christian(1992): Uempire des cartes.Approche thorique de la cartographie travers
Phistoire,Albin Michel, Paris.
Latour,Bruno (1987): ScienceIn Action. How to Follow Scientistsand EngineersthroughSocie-
ty,HarvardUniversityPress,Cambridge Mass.
Latour, Bruno (1988 a): The Pasteurizationof France, Harvard UniversityPress, Cambridge
Mass.
Latour, Bruno (1988 b): The Politics of Explanation: An Alternative,in: S. Woolgar (Hrsg.),
Knowledgeand Reflexivity. New Frontiersin the Sociology of Knowledge,Sage,London, S.
155-177.

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions
On actor-network
theory 381

Latour,Bruno (1988 c): A RelativistAccount of Einstein'sRelativity,Social Studiesof Science

18, S. 3-44.
Latour,Bruno (1994 a): On Technical Mediation, Common Knowledge 3(2), S. 29-64.
Latour, Bruno (1994 b): Une sociologie sans objet? Note thorique sur Pinterobjectivit,
Sociologie du travail36, S. 587-607.
Latour,Bruno (1996 a): Flat-Earthersand Social Theory,in: M. Power(Hrsg.), Accountingand
Science: Natural Inquiryand Commercial Reason, CambridgeUniversityPress,Cambridge
Mass., S. xi-xviii.
Latour,Bruno (1996 b): Do scientificObjects Have a History?Pasteurand Whiteheadin a Bath
of Lactic Acid, Common Knowledge 5(1), S. 76-91.
Latour, Bruno (1996): Der Berliner Schlssel. Erkundungen eines Liebhabers der
Akademie Verlag,Berlin (bersetztvon Gustav Roler).
Wissenschaften,
Lee, Nick, Brown,Steve(1994): Othernessand the Actor-Network:The UndiscoveredConti-
nent,AmericanBehavioralScientist37, S. 772-790.
Lynch,Michael (1985): Art and Artifactin LaboratoryScience.A Studyof Shop Workand Shop
Talk in a ResearchLaboratory,Routledge,London.
Mol, Annemarie, Law, John (1994): Regions, Networks, and Fluids: Anaemia and Social
Topology,Social Studiesof Science, 24, S. 641-672.
Prigogine,Ilya, Stengers,Isabelle (1979): La nouvelle alliance, mtamorphosede la science,
Gallimard/Bantam,Paris/New York.
Serres,Michel (1983): Hermes. LiteratureScience Philosophy,The John Hopkins University
Press,Baltimore.
Serres,Michel (1987): Statues,Franois Bourin, Paris.
Serres,Michel (1995): Conversationson Science, Culture and Time with Bruno Latour, The
Universityof Michigan Press,Ann Arbor.
Stengers,Isabelle (1993): L'inventiondes sciences modernes,La Dcouverte,Paris.

This content downloaded from 138.251.236.241 on Fri, 6 Feb 2015 04:38:42 AM

All use subject to JSTOR Terms and Conditions