Norw. Arch. Rev., Vol. 22, No. 2, 1989

The Bordes Method?E. M. KOLPAKOV and L. B. VISHNYATSKY

University of Leningrad, USSR

Institute of Archaeology, Leningrad, USSR

Up until the present the Bordes method has been widely accepted amongresearchers of the palaeolithic throughout the world. However, its relevanceto the tasks for which it is usually applied has not been proved and, more-over, is very doubtful. Both typologically and statistically the method ispernicious. It is necessary to elaborate a new approach to the analysing ofstone industries on the grounds of realization that there can be and must bemore than one classification for the same material.

Methods are tools. They are developed forspecific tasks . . . But methods, like tools, can beabused. The most oblivious form of abuseinvolves using methods not because they fit thetask in hand, but because they are methods weknow and can easily apply (Moore & Keene1983:4).

. . . only when current theories and methodshave been fully evaluated and their weaknessesexposed can new theories effectively take theirplace (Collins 1970:17).

In the forty or so years of its existence theBordes method has, on more than oneoccasion, been a subject of debate amongpalaeolithic archaeologists. Individualaspects of it have been elaborated and addedto and have sometimes been subjected tomore or less severe criticism (Mason1967:758-759, Kerrien & Clark 1968, Cahen& van Noten 1970, Kozlowski 1972:455-457,Suleimanov 1972:74-76, Matyushin 1975:44,Gladilin 1976:4-27, Praslov 1984:100-101,Minzoni-Déroche 1985, and so on). On thewhole, however, this method is still pre-dominant in palaeolithic studies, and theassociated path for the analysis of archaeo-logical material has even been referred to asthe principal 'highway'.

In this article, which consists basically ofan analysis and evaluation of the Bordesmethod as a whole and of its componentparts, we have attempted to show that in factBordes' 'highway' leads up a blind alley, andthat it is necessary to adopt another method

of analysing material, based on differentprinciples.

We will first try to answer the apparentlysimple question 'what does the Bordesmethod consist of?'. We will use for this the(in our view, exhaustive) formulation of themethod given by D. Sonneville-Bordes. Bylooking at his works, it is easy to ascertainthat this formulation fully corresponds to themore extensive description of the method byBordes himself (Bordes 1950, 1961a, 1984and others). Thus the Bordes methodrequires:

(1) the establishment of a typology which recog-nizes, defines and describes types throughtechnical and morphological analysis and,possibly, classifies them in groups;

(2) the choice of a type-list which orders thesetypes and groups. An inventory, with per-centages, which records all the stratigraph-ically well-defined tools, excluding rawmaterial, which is analysed separately;

(3) cumulative graphs which, combining the fre-quencies of tools by type, allow a graphic rep-resentation of the dynamic or trend of theseries;

(4) indices and characteristic groups, resemblingone or several technically, typologicalJy or cul-turally significant types which may be isolatedand represented by block diagrams (Sonn-eville-Bordes 1974-75:9-10).

So, it appears, the question is clear and theessence of the method fully understood. Letus, however, turn our attention to the first

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

108 E. M. Kolpakoo and L. B. Vishnyatsky

point, where it is stated that the methodrequires 'the establishment of a typology'.This is a key point: the success — and thevery possibility — of implementing all thesubsequent procedures depends on the qual-ity of the typology. The question is, howshould this typology be established?

Unfortunately, practically nothing is saidon this score, and no instructions as tomethod are given, either in Bordes' works orin those of his followers. The procedure forsetting up a typology is represented as beingself-evident, not requiring any kind of expla-nation or recommendation as to method(unless, of course, Bordes' references to a'typological eye', acquired as a result ofmuch experience of classifying, are to betaken as such).1 It is quite clear that withouta typology (a type-list) the Bordes method isinconceivable, but, at the same time, it doesnot provide a method for the construction ofa typology. Bordes created his own typology,but paid no attention to the question of howhe did this and how, in general, to createoriginal typologies. So our first conclusion isthat the Bordes method is not a method forconstructing a typology.

Nowhere in Bordes' works does he statedirectly that his own type-list must be appliedfor the analysis of different assemblages fromdifferent regions. On the contrary, the factthat formulations of the method (see above)are concerned with the creation of type-listscan be regarded as an indication of the neces-sity of constructing a special one for each sep-arate case. Nevertheless, the method itselfmanifests a very strong tendency towards apractical change in these procedural steps —towards the use of one and only one type-list.The fact is that all the subsequent operationsmake sense only if material from differentsites is processed according to one type-list.Only then is it possible to compare cumu-lative graphs and indices (and any other par-ameters based on percentages) and to drawconclusions with their help. The more sitesthat are processed according to one type-list,the more effective is the method. It is no acci-

dent that Bordes himself, as far back as themid-1950s, began to apply his own type-list,based on Mousterian material fromsouthwest France, not only to material fromFrance and Western Europe, but also fromthe Near East, North Africa and so on.

Thus two different interpreations of theessence of the Bordes method are equallypossible. According to one of them, atypology should be created for a particularaggregate of material and then, on the basisof this, the prescribed statistical operationsshould be carried out; according to the otherinterpretation, the material should be classi-fied in accordance with Bordes' type-list andthe statistical processing should be carriedout on this basis. We emphasize that thepossibility of making either interpretation isinherent in Bordes' own work.

Both possibilities mentioned have beenrealized in archaeological practice. It is truethat the Bordes method has been perceivedas a sequence of procedures — the first ofwhich is the construction of an originaltypology — only by specialists in the UpperPalaeolithic (Sonneville-Bordes & Perrot1954, 1955, 1956, Tixier 1967, Abramova1979a, 1979b), where any other inter-pretation of it is clearly not possible.

Moreover, Sonneville-Bordes andPerrot's original type-list came to beregarded by some as universal for the UpperPalaeolithic (Chernysh 1959, 1967). Thisinterpretation of the Bordes method, inwhich the principal place is occupied by histype-list, undoubtedly predominates amongpalaeolithic archaeologists.

For most specialists, the application of themethod under examination consists, first ofall, in the division of their assemblagesaccording to the classic type-list, which, inthis way, is in practice perceived and used asa universal skeleton-key to any material.. Itis true that, more than 20 years ago, duringthe years of the triumphal march of the Bor-des method through the palaeolithic assem-blages of the world, some of the researcherswho had enthusiastically propagandized it

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

The Bordes Method? 109

wrote that a 'deeper study of flint artefactsnecessitates the creation of new regionaltypological lists, more detailed for particularterritories', but even in this case, Bordes'type-list itself was considered to be the basisof such lists, 'the framework for future pro-cessing' (Lyubin 1965:74-75).

If we turn to the publications of the pastfifteen years, it is easy to see that, forresearchers using the Bordes method, devi-ating from the classic type-list when settingup typologies is virtually impermissible.Only the occasional addition of new types tothe list passes for the creative developmentof the method, along with the frequentchanges introduced in the system of indices.There are, of course, examples where realattempts are made, on the basis of the classictype-list, to create fuller 'regional typologicallists' (e.g. Kolosov 1986:16-18, Lyubin1977), but in the majority of cases assem-blages are simply assigned to the 63 cat-egories on Bordes' list, or, more precisely,to those categories which manage to furnisha more or less suitable correspondence.

We will give some examples to illustratewhat has been said. Piperno found 36 cat-egories from Bordes' list sufficient for theJahrom (Iran) assemblage, after which all hehad left was a single pièce esquille (Piperno1972). Akazawa also allotted the materialsfrom level D of Shanidar to 36 categoriesfrom the classic type-list, adding only 'retou-ched rods' and declaring, moreover, that hewas using 'a modified version of Bordes'type-list' (Akazawa 1975:5).

A. K. Dzhafarov, analysing material fromthe Talgarskaya cave (Azerbaijan), man-aged with 20 categories in the final table(1983:58), which swallowed up the assem-blage without a trace. For the first layer ofShaitan-Koba (Crimea), 32 categories wereenough (Kolosov 1972:24); for the Ketrosasite (Soviet Moldavia), 33 categories (Anis-yutkin 1981:48); and for Khudgi (Tadjiki-stan) 34 (Ranov & Amosova 1984:30-31).Similar examples could be cited ad infinitum.

Applying Bordes' type-list so widely and,

so to speak, automatically — using it essen-tially as a determinant — would be fully jus-tified and in order only if the diversity of ourmaterials always corresponded to the diver-sity of the material on which the classic type-list is constructed, and if the fixed com-binations of attributes which composed Bor-des' types were fixed for all complexes. It isclear, however, that in any new materialother combinations of attributes may turnout to be fixed. If we then impose the classictype-list on them we will obtain a picturewhich does not reveal the material's charac-teristic structuredness or separate what isinherent in the material from what is acci-dental — the primary and basic task of ourtypology — but, on the contrary, which onlyconceals it.

The objection could be raised that all thearguments put forward here are of a veryabstract nature, and Bordes' type-list has,moreover, apparently shown its viability inpractice. Until now, the most varied assem-blages from the most varied regions havebeen successfully processed with its help. Ithas been used hundreds of times. Has it,however, been verified?

Is it.really possible to consider that thosehundreds of occasions when material from agreat many sites from different regions hasbeen processed according to the classic type-list constitute a verification of it? This type-list is a priori in relation to every newly stud-ied complex and, before applying it, it isworth asking whether, in the case in ques-tion, another typology, not provided for byBordes' type-list, might perhaps not be moresuitable for the material. To check this, it isnecessary, laying aside the ready-madedeterminant and, as far as possible, ignoringit, to try to construct a typology stemmingonly form a particular aggregate of materialsbelonging to one complex and, of course,one culture. This approach is obviously morecorrect than immediately and mechanicallytransferring onto the new material the struc-ture revealed in complexes which are geo-graphically very distant and which, more

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

110 E. M. Kolpakov and L. B. Vishnyatsky

than likely, belong to different cultures,2

thereby thrusting on it what is most likely analien system of attribute-linking. As far aswe know, however, none of the researchersusing the Bordes method has reported anysuch check being conducted.

Therefore, when applying Bordes'typology a priori, we are probably very oftenjust squeezing material into a frameworkprescribed for it and creating the illusion thatthe material corresponds to Bordes'typology. The character of the definition ofmost of the types in Bordes' list is such thatit is, as a rule, easy to do this, whereas itis difficult or impossible to notice a strainedinterpretation from drawings of a few selec-ted standard objects. In theory, it is possibleto divide any material in accordance with theBordes method, but it by no means followsthat it is necessary to do so, or that such adivision is the only possible and correct one.

In addition, work with data obtained bydifferent authors using a single type-list(Bordes') presents quite a serious problem.The fact that two researchers have the sametypological system is no guarantee that,when working with the same material, thefinal 'haul' in each of the cells of this systemwill turn out to be the same. To put it moresimply, if two researchers are set to workwith one set of material and one type-list theywill consequently (at least in many cases)obtain different results. There are data fromexperiments of this kind but, unfortunately,as far as we know, they have not been pub-lished. Even so, there is fortunately somematerial in the literature that allows tosubstantiate our claims.

In 1984 Dibble published the results of hisresearch on the Mousterian material fromthe Bisitun cave in Iran, which was pre-viously studied by Skinner (1965). Bothauthors gave a typology in accordance withBordes' type-list and used his definitions andindices. The differences between the totalfigures turned out to be appallingly great.Where the number of tools of a particulartype given by Dibble exceeds that given by

Skinner this can still sometimes (though byno means always) be explained by the factthat Dibble had access to materials fromlevel G, which were not studied by Skinner.But for 10 types the position is reversed andthe discrepancies are not in ones but in tens(Nos. 6, 7, 21, 31 and so on). Finally,although Skinner discovered equivalents ofonly 24 types from Bordes' 62-category listin the Bisitun assemblage, Dibble used 47items; that is, twice as many.

In order to really decide whether Bordes'typology is adequate for various materialsand whether his definitions of types are satis-factory for different cases, a special checkand study are necessary. Such a check isimpossible without completing new classifi-cations of material directly according toassemblage. At present it is impossible toavoid the conclusion that the tradition whichhas arisen of using Bordes' type-list has nofactual or methodological basis and, mostprobably, distorts our understanding of theoriginal material.

We now turn to the other interpretation ofthe Bordes method, which has foundexpression in several studies of the UpperPalaeolithic of Europe, North Africa andSiberia. Here, too, a whole number of prob-lems arise, the solutions to which cannot befound in the work of Bordes or his followers.

First, as we have already noted, it is uncer-tain how we can establish a typology andregulate it (construct a type-list) in an alto-gether methodologically correct way.

Secondly, for the construction of atypology the aggregate of materials withinwhose boundaries it is created is very impor-tant. Even with an intuitive division, typeswill take shape through some replicability ofcombinations of attributes, considered to becharacteristic and significant. Here thedegrees of replicability and fixing of differentcharacterisics should be assessed, and,through them, the degree of similaritybetween artefacts. But this makes sense onlywithin some form of limited aggregate ofdata, in so far as, with the broadening or nar-

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

The Bordes Method? I l l

rowing of its parameters (if the material isnot absolutely homogeneous), other com-binations of attributes may turn out to rep-licate themselves more in new parameters,and it may be possible to evaluate the degreeof similarity between artefacts in a differentway. Consequently, changes in the typologywill depend on changes in the parameters ofthe aggregate within which it is formulated.The problem inevitably arises of assessingthe criteria for choosing the aggregate to betypologized, and of defining its parameters.

Thirdly, the question arises of determiningwhen the existing type-list can be applied tonew material and when it cannot. As yet thisis successful only for clear-cut cases. For theoverwhelming majority it is possible to dothis only after constructing a separatetypology for the new group of material whichinterests us. If it differs to some extent fromthe existing typology, then which of themshould be used for research? Or is it nec-essary to construct a general type-list for anew assemblage which includes old and newmaterials?

Fourthly, the problem of working withdata inserted in different type-lists remainsmethodologically unresolved. Assemblagesentered in different type-lists should be com-pared outside the confines of the method,that is, without applying its statistical pro-cedures. And, in fact, in this interpretationof the Bordes' method, only the statisticalpart is used when original typologies areconstructed — for the actual process of theirconstruction is in no way regulated by themethod.

We now turn to an examination of the stat-istical part of the Bordes method. We recallthat it can be reduced to two basic methods:

1. The expression in percentages (indices) ofthe representation in complexes of certaintypes, groups of types or technicalelements.

2. The graphic expression of relationshipsbetween complexes (cumulative graphs,histograms).

These methods are evidently extremelyelementary, and they came into use in StoneAge archaeology before the Bordes methodappeared. Bordes only combined them intoa single research process, and suggested anumber of indices not previously used.

Percentages and cumulative graphs are,however, applied incorrectly in the Bordesmethod. They must not be used without cal-culating confidence intervals, otherwiseserious mistakes are possible even at thedata-processing stage. Attention has beenpaid to this in the archaeological literature(e.g. Pislary & Pozhidaev 1982). Even moreimportant is the fact that the interpretivepossibilities for cumulative graphs are toostrongly dependent on the order in which thetypes occur in the type-list. Kerrich & Clarkhave demonstrated this in a specificallyarchaeological context (1968:57-64).Whether cumulative curves constructed outof exactly the same data will appear similaror dissimilar depends on the sequence oftypes in the type-list. The same material canbe put in order and entered in graphs dif-ferently, and in this way it is easy to supportany intuitive method by the correspondingconstruction of a cumulative graph.

In this way, the 'typologically statistical'(as it is sometimes called: e.g. Vekilova1971:120) Bordes method turns out to beunsatisfactory as regards both typology andstatistics.

But, the objection might again be raised,was it not precisely the application of theBordes method which allowed importantconclusions about the technical and tech-nological heterogeneity of the Mousterian(which many interpret in the sense of theexistence of different cultural traditions) tobe drawn? Does this not mean that themethod works after all?

We will examine the Bordes method in itsoriginal application on French material. Sev-eral types of Mousterian were isolated: typi-cal, denticulate, Charentian and from theAcheulian tradition. The characteristics ofeach of these are well known and there is

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

112 E. M. Kolpakov and L. B. Vishnyatsky

Fig. 1.

no need to repeat them here. Bordes wasincluded to interpret these types of Mou-sterian as different archaeological cultures(e.g. Bordes 1960:102-103, 1961b:807, Bor-des & Sonneville-Bordes 1970:64).

We will start with the cumulative graphs.We have transferred all the cumulativecurves on Mousterian sites cited by Bordes1984) onto one graph (Fig. 1). It is apparentfrom this that in fact the groups of curvesdefined by Bordes and interpreted as charac-

teristic of different cultures hardly exist. Thecurves are disposed more or less evenly onthe field of the graph, between the extremevalues.3 The percentage indices of types forall the sites presented by Bordes changesmoothly, without forming statisticallysignificant clusters. This was also reflected inthe values of the indices: they also changesmoothy — a concentration can. be observedonly in the area of the zero values. And Bor-des even used standard complexes, which

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

The Bordes Method? 113

must be characteristic of Mousterian cul-tures. If to this is added the existence of com-plexes occupying an intermediate positionand confidence intervals, which in this caseare mandatory and which, nevertheless,have not been studied, and if Kerrich &Clark's instructions are borne in mind, thenit becomes clear that statistically significantdifferences between the Mousterian culturesdefined by Bordes are not observed or, inany event, their existence can by no meansbe considered proved or demonstrated.

It is also very characteristic that a total ofonly four categories of tool (denticulate,bifaces, scrapers and backed knives) are con-sidered important and used for distinguish-ing particular types of Mousterian, and onlytwo or three for each compared pair of types.Moreover, the internal differentiation ofthese types on the type-list is practicallyinsignificant and means — and this should beemphasized—that the type-list itself is of nosignificance. This will probably seem para-doxical, but it is the case. Cumulative graphsconstructed according to type-lists do not tellus anything, and the way that they are usedcannot tell us anything. In fact it turns outthat it is not the Bordes method at work, butsomething else.

We emphasize that what we have saidabout the method in no way excludes thepossibility that those types of Mousteriandefined by Bordes exist in France. They mayeven have been correctly defined, but this isnot thanks to the Bordes method.

We would like to comment further on thetypes of Mousterian defined by Bordes(possibly digressing slightly from ourimmediate theme). They have begun to playthe same role in palaeolithic archaeology astool-types from the classic type-list, that is,they are in fact often used as a kind of uni-versal type-list of cultures. This was, to asignificant extent, facilitated by the work ofBordes himself, who wrote that whileresearching European, North African andMiddle Eastern complexes he had observed'several recurrent types of Mousterian, more

or less well represented in these differentregions', accompanying this claim with thefamous list (Bordes 1977:37-38).

Armed with Bordes' type-lists, Sovietresearchers also got down to work. "Theypicked out from their materials the samevariants of the Mousterian flint industry asin France' (Formozov 1977:33). In this way,mysterious 'paths' and 'lines' of developmentappeared, undoubtedly originating fromBordes' 'type-list of cultures'4 and resultingfrom its imposition on the original material.In our view, the correct explanation of suchstartingly apparent like-mindedness of theinhabitants of the Mousterian oikumen wasgiven by Praslov. He remarked that 'the pro-cessing of material according to one schemeand a single set of criteria, suggested by Bor-des, always leads to the results envisaged bythis system. It is for precisely this reason. . .that single 'paths' or 'lines' of developmentof Mousterian cultures in different territoriesarise' (Praslov 1984:101).

It is interesting that the present situationwith 'types of Mousterian' or 'paths of devel-opment' is very reminiscent of that whichpertained around thirty years ago in thearchaeology of the Upper Palaeolithic, whenit was in the throes of separation from thestandard French scheme; true, not a culture-dividing scheme but a periodizing one —Aurignacian, Solutrean and Magdalenian.Archaeologists, most of whom have had ahistorical education, could, it seems, learn alesson from the history of their own disci-pline, which testifies that our (unfortunatelysilent) material can be squeezed at will intothe Procrustean bed of any scheme, howeverunsuitable.

We will now briefly formulate our con-clusions about the essence and quality of theBordes method:

1. There are two possible interpretations ofthe essence of the method and two poss-ible approaches to it. It can be seen eitheras simply a succession of distinct researchoperations (creation of an original

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

114 E. M. Kolpakov and L. B. Vishnyatsky

typology; calculation of percentages; con-struction and visual comparison of graphsand histograms), or as primarily a methodof working based on a universal classictype-list. The second approach to themethod is the predominant one.

2. In any interpretation of the essence of theBordes method it is (a) not a method forconstructing a typology, and (b) primitiveand basically unacceptable in its statisticalaspects.

3. The use of the Bordes method as a uni-versal basis for typologizing is essentiallyincorrect, not well founded in practice andtherefore unacceptable. In fact, it violatesthe material, thrusting upon it a pre-ordained structure rather than revealingits own particular one.

4. The results obtained on the basis of theBordes method need to be verified and theidea that the types of Mousterian definedby Bordes are a 'type-list of cultures' mustbe eliminated.

Having thus formulated our conclusionswe could draw a line here and consider ourtask completed. However, it is precisely inconnection with the above conclusions thatat least two more questions arise, an exam-ination of which — albeit a short one —seems to us unavoidable.

The first question is: if the Bordes methodis so bad, what then is the secret of its dura-bility and why is it still popular?

The second question is: if the Bordesmethod is no good, what should take itsplace?

THE SECRET OF THE BORDESMETHOD'S SUCCESS

The appearance of the Bordes method at thestart of the 1950s was undoubtedly a signifi-cant step forward in the development of pal-aeolithic archaeology. It replaced the'typological muddle' (as Lyubin (1965:14)aptly expressed it) and extreme subjectivity

of the evaluation of similarity and non-simi-larity. In place of terminological and con-ceptual chaos came order (standardization ofterminology). In place of the judgement oftypes as 'more similar' or 'less similar' cameindices, graphs and histograms (filling theneophytes from the humanities with holydread). A single typological foundationapparently paved the way for statistics, andstatistics (even elementary statistics), apartfrom anything else, satisfied the archae-ologists' desire that their discipline shouldbecome more scientific, and answered thedemands of the time and of fashion. All thisensured an interested reception for themethod among the majority of specialists,and its consequent victorious march throughthe palaeolithic assemblages of manycountries of the world. The Bordes methodowes its importance in palaeolithic archae-ology and its great influence on the minds ofresearchers not, of course, to the methods inthemselves, or to the statistical techniques,but, primarily, to their combination with adetailed typology which has (largely incor-rectly) been taken as universal.

This combination created an illusion of theobjectivity in comparisons and conclusionstowards which palaeolithic archaeologistshad long been striving. And when it gradu-ally became clear that a universal typologyfor the study of culturally and historicallyspecific complexes and their cultural cor-relation was simply not possible, it turnedout that giving up the hard-won illusion wasfar from simple.

The method survived the criticism and isagain popular among palaeolithic archae-ologists.

It is popular for a number of subjectivereasons.

It is popular because the nature of thetypes from the classic type-list allows somekind of correspondence to be selected forthem in a very wide range of complexes.5

It is also popular because there appears tobe no valid alternative, although the variousattempts to create one are well known.

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

The Bordes Method? 115

WHAT IS THE SOLUTION?Specialists who find the Bordes methodunsatisfactory sometimes put forward theirown systems for analysing stone artefacts. Intheir typological aspects, however, these sys-tems usually not only repeat the basic short-coming of the Bordes method — that is, theclaim to universality of typology — but alsoadd to it another equally real defect: a prioritypology. The authors of such constructionsseem to think that it is sufficient to observethe logical rules for a division into rankedconceptual categories which produce anordered hierarchy of taxa—and all the prob-lems will be solved.

Klejn, in his seminal monograph 'Archae-ological typology', demonstrated the falsenessof the path leading to 'a priori universal classi-fications-typologies of Gorodtsov's canoni-cal type' (Klejn 1982:esp. 266-267). Manypeople are evidently won over by the close-ness of such classifications to Linnaeanclassification — considered to be the ideal —and by the possibility of dividing materialaccording to a precise 'previously definedscheme' (Gladilin 1976:29). We repeat, how-ever, that before thrusting some scheme onthe material, it is worth asking whether ithas its own inherent scheme. Or, as biol-ogists would express it, 'the form of a systemmust be revealed rather than postulated'(Lyubischev 1923, see also Meien 1978;503,Chaikovsky 1986:53).

Yes, a hierarchical classificatory systembased on formal logical bases can be useful,but only for a definite purpose, specificallyfor preliminary auxiliary ordering ofmaterial.

The current practice in archaeology is tomake a single classification of one set ofmaterial. However, the view that it is per-missible, or even essential, to allow the cre-ation of different classifications for the samematerial, has been around for some time.6

This primarily concerns the division ofclassifications into auxiliary and research(Gryaznov 1969, Kamenetsky et al. 1975,Klejn 1979, Klejn 1982).

Auxiliary classification is intended forinitial ordering and description of material;for quick and reliable searches of the nec-essary data for various studies; for theinclusion of extensive materials in a conciseform accessible to every researcher; and soon. Such a classification must permit theimmediate entry of newly discoveredmaterial and, at the same time, reflect themass of characteristics of the objects whichare necessary for the inclusion of thismaterial in research. In order to turn toresearch it is usually necessary to somehowinclude extensive material, to get a classi-ficatory idea of it, to select from it what ispresumed to be necessary and so on. Onlyafter this does the research itself, includingthe creation of research classifications,become possible. Auxiliary classification,like research classification, is intended toserve research tasks, but mainly at the pre-liminary stage. This is the root of their dif-ferences. Auxiliary classification must there-fore make a preliminary specification of theset of attributes and their possible meanings;the attributes must be relatively easilyacceptable, unambiguous and quickly de-finable; the structure of the classificationmust be created beforehand and be usefulin practice; all terminology must bestandardized and simple.

Research classifications aim not just atordering and description of material, but atthe revelation of some kind of essential linksand characteristics in it. They and theirresults are better termed 'typology'.

The most obvious and usual aim ofarachaeological research is the uncovering ofthe cultural specifics of a complex and its cor-relation with other complexes. When con-structing such a typology it is possible toproceed solely from the cultural context ofthe given complex, or of several complexesif they are known to belong to the same cul-ture. Each complex then begins to generateits own type-list, but the limits of the uni-versality of a type-list are the framework ofone archaeological culture.

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

116 E. M. Kolpakov and L. B. Vishnyatsky

Of course, even with this approach, amultitude of problems arise and demand tobe solved — first of all, the problem of com-paring the different authors' data. As it is notpossible to examine this in detail here, wewill only note that the traditional approachalso — as shown by the example of Skinnerand Dibble's typologies for Bisitun — in noway rescues us from the necessity of takingaccount of and overcoming the subjectivityof our typological values.

Bordes' typology fulfils both the auxiliaryand the research functions (as does Sonn-eville-Bordes and Perrot's typology),although in fact it was originally intendedchiefly for the research function and for fullydefined material. The failure to distinguishbetween these classifactory systems has hadfar-reaching negative consequences. Des-cription, reporting and publication of findsare conducted according to Bordes' type-list, and all research is also carried outon exactly the same basis, as recommendedby Bordes himself. Bordes' typology,however, is not suitable per se for studyingnew material — as indicated in the main partof our article. It is also poorly suited to therole of auxiliary classification. In Bordes'works there is neither a system of attributesfor description, nor the means of establishingthem for new materials.

The fact that work based on Bordes'typology began to acquire automatically thestatus of scientific research for this reasonalone, has long played an extremely negativerole; the application of Bordes', or anyother, typology in research does not in anyway ensure that it is scientific. The use ofhis typology in description and publication ofmaterial does not in itself ensure the qualityand scientific nature of this work. Thequicker we rid ourselves and the literatureof the illusion that Bordes' types alone arescientific, the less we will have to alter andthe sooner we will obtain new scientific infor-mation.

The question of the necessity of carryingout both auxiliary and research classifi-

cations and of the impossibility, as a rule,of combining these two functions in a singleclassification has ceased to be a purely theor-etical question in palaeolithic archaeology. Itis now a practical question: wide experienceof work with many kinds of material hasalready shown that we cannot get by withoutcreating an auxiliary classification, separatefrom the research classification, if we do notwant our discipline to stand still. This meansthat the Bordes method must be replaced notby a new typology and some kind of methodbased on it, but by an auxiliary classificationintended especially for it, and by separateresearch classifications together withmethods for constructing them and workingwith them.

NOTES

1 'One has to see a great number ofimplements, classify them, see them againseveral times, before one acquires a "typo-logical eye"' (Bordes 1972:141).

2 Bordes' type-list is itself quite eclectic fromthis point of view, being only a guide to'the definition and illustration of types',combined with a single list based on thechoices of a single researcher. As Par-kington, in our view quite correctly, noted(although not specifically in relation toBordes' type-list), 'many of the implementand cultural type names were coined in thenineteenth or early twentieth centurieswhen they proved useful reference pointsin a discipline still groping in the dark.However, there was a danger that thesereference points would become pigeon-holes into which further data would bepushed or at times forced' (Parkington1972:11, quoted in Klejn 1982:91).

3 'Given the same pair of curves someobservers might find them to be quite simi-lar and others might find them quite dis-similar'. This is one of the particularweaknesses of the method, as noted byKerrich & Clark (1958:68-69; see furtherMinzoni-Déroche 1985).

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

The Bordes Method? 117

4 For example, in the Caucasus, Lyubinentered all the Mousterian cultures hedefined into three lines of development:typical, denticulate and Charentian (Lyu-bin 1977, 1984:69-70).

5 Such a possibility, latent in the type-list, isthe sort of trap which it is possible to escapefrom only once one has understood its con-struction, which is far from obvious.

6 A similar necessity is also felt in manyother disciplines, in particular, in biologywhere 'since Linnaeus' time a debate hasbeen going on about the fact that at leasttwo systems are necessary — an artificialone, useful for locating a species, and anatural one, expressing, as Linnaeus said,the essence of the thing (bearing in mindthat naturalness is itself a relative concept)'(Chaikovsky 1986:48).

REFERENCES

Abramova, Z. A. 1979a. Paleolit Eniseya. Afon-tovskaya kultura. Novosibirsk.

Abramova, Z. A. 1979b. Paleolit Eniseya. Koko-revskaya kultura. Novosibirsk.

Akazawa, T. 1975. Preliminary notes on themiddle palaeolithic assemblage from Shanidarcave. Sumer 31(1), 3-10.

Anisyutkin, N. K. 1981. Arkheologicheskoeizuchenie must'erskoi stoyanki Ketrosy.Ketrosy, Moscow.

Bordes, F. 1950. Principes d'une méthode d'étudedes techniques et de la typologie du paléo-lithique ancien et moyen. L'anthropologie 54.

Bordes, F. 1960. Evolution in the paleolithic cul-tures. In Evolution after Darwin 2 (The evol-ution of Man). Chicago.

Bordes, F. 1961a. Typologie du paléolithiqueancien et moyen. Bordeaux.

Bordes, F. 1961b. Moustérian cultures in France.Science 134

Bordes, F. 1972. On old and new concepts intypology. Reply. Current Anthropology 13(1).

Bordes, F. 1977. Time and space limits of the mou-sterian-stone tool as cultural markers. Canberra

Bordes 1984. Leçons sur le paléolithique 2 (lepaléolithique en Europe). Paris.

Bordes, F. & Sonneville-Bordes, D. 1970. The

significance of variability in palaeolithic assem-blages. World Archaeology 2(1).

Cahen, D. & Van Noten, F. 1970. 'Stone agetypology: another approach. Current Anthro-pology 12(2).

Chaikovsky, Yu. V. 1986. Grammatika biologiya,Vestnik AN SSSR 3.

Chernysh, A. P. 1959. Pozdnii paleolit SrednegoPridnestrov'ya. Paleolit Srednego Pridnes-trov'ya. Moscow.

Chernysh, A. P. 1967. O nomenklature pozdne-paleoliticheskikhorudii. Kratkiesoobshcheniyainstituta arkheologii 3.

Collins, G. 1970. Stone artefact analysis and rec-ognition of culture traditions. World Archae-ology 2(1).

Dibble, A. 1984. The Mousterian industry fromBisitun Cave (Iran). Paleo-Orient 10(2).

Dzhafarov, A. K. 1983. Must'erskaya kul'turaAzerbaidzhana. Baku.

Formozov, A. A. 1977. Problemy etnokul'turnoiistorii kamennogo veka na territorii evropeiskoichasti SSSR. Moscow.

Gladilin, V. N. 1976. Problemy rannego paleolit aVostochnoi Evropi. Kiev.

Gryaznov, M. P. 1969. Klassifikatsiya, tip,kul'tura, Teoreticheskie osnovy sovetskoi arkh-eologii. Leningrad.

Kamenetsky, I. S., Marshhak Z. E. & Sher, Ya.A. 1975. Analiz arkheologicheskikh istochini-kov (Vozmozhnosti formalizovannogo podkh-oda). Moscow.

Kerrien, J. & Clark, D. 1968. Notes on the poss-ible misuse and errors of cumulative percentagefrequency graphs for the comparison of pre-historic artefact assemblages. Proceedings ofPrehistoric Society 33.

Klejn, L. S. 1979. Ponyatie tipa v sovremennoiarkheologii. Tipy v kul'ture. Leningrad.

Klejn, L. S. 1982. Archaeological Typology.Oxford.

Kolosov, Yu. G. 1972. Shaitan-Koba —must'erskaya stoyanka Krimu. Kiev.

Kolosov, Yu. G. 1986. Akkaisakaya must'ershayaukl'tura. Kiev.

Kozlowski, J. 1972. The typological classificationof stone artefacts (contribution to discussion).Sprawozdania Archeologiczne 24.

Lyubin, V. P. 1965. K voprosu o metodike izuch-eniya nizhnepaleoliticheskikh kamennykh oru-dii. Materialy i issledovaniya po arkheologiiSSSR 131. Moscow-Leningrad.

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4

118 E. M. Kolpakov and L. B. Vishnyatsky

Lyubin, V. P. 1977. Must'erskie kul'tury Kav-kaza. Leningrad.

Lyubin, V. P. 1984. Rannii paleolit Kavkaza.Paleolit SSSR. Moscow.

Lyubishchev, A. A. 1923. O forme estestvennoisistemi organizmov. Izv. Biol. Nil pri Perm.un-te 2(3).

Mason, R. 1967. Analytical procedures in theearlier and middle stone age cultures in Sou-thern Africa. Background to Evolution inAfrica.

Matyushin, G. N. 1975. O metodakh klassifikatsiimassovogo inventarya pamyatnikov kamen-nogo veka. Istoriya material'noi kul'turyUzbekistana 12.

Meien, S. V. 1978. Osnovnye aspekty tipologiiorganizmov. Zhurnal obschei biologii 39(4).

Minzoni-Déroche, A. 1985. Lithic artefacts inter-pretation: an empirical approach. WorldArchaeology 17(1).

Moore, J. & Keene, A. 1983. ArchaeologicalHammers and Theories. New York.

Parkington, J. 1972. Stone implements as infor-mation. The Interpretation of ArchaeologicalEvidence. Claremont Cape.

Piperno, M. 1972. Jahrom, a middle palaeolithicsite in Fars, Iran, East and West 22(3-4).

Pisiary. I. A. & Pozhidaev, V. F. 1982. O pri-menenii metoda protsentnykh sootnoshenii.

Metodologicheskie i metodicheskie voprosiarkheologii. Kiev.

Praslov, N. D. 1984. Rannii paleolit Russkoi rav-niny i Kryma. Paleolit SSSR. Moscow.

Ranov, V. A. & Amosova, A. G. 1984. Raskopkimust'erskoi stoyanki Khudzhi v 1978g. Arkh-eologicheskie raboty v Tadzhikistane, 1978 18.Dushanbe.

Skinner, J. 1965. The Flake Industries ofSouthwest Asia: A Typological Study Ph. D. dis-sertation, Columbia University.

Sonneville-Bordes, D. 1974-75. Les listes types.Observations de méthode. Quaternaria 17.

Sonneville-Bordes, D. & Perrot, J. 1954-56 Lexi-que typologique du paléolithique supérieur.Bulletin de la Société Préhistorique Française,51-53.

Suleimanov, R. Kh. 1972. Statisticheskoe izuch-enie kul'tury grata Obi-Rakhmat. Tashkent.

Tixier, J. 1967. Procédés d'analyse et questions determinologie concernant l'étude des ensemblesindustriels du paléolithique récent et de l'épi-paléolithique dans l'Afrique du Nord-Ouest.Background to Evolution in Africa.

Vekilova. E. A. 1971. Kamenny vek Kryma.Nekotorye itogi i problemy. Materialy i issle-dovaniya po arkheologii SSSR 173. Leningrad.

Translation by Sarah Wright

Dow

nloa

ded

by [

] a

t 05:

38 0

1 M

arch

201

4