Dokumentationsproblemer indenfor arkæologien

Resumé

Problems of documentation in archaeology

When documentation is spoken of in connection with archaeology, it is usually in the sense of the presentation of proof for the interpretation of observations made during excavation or during the study of artefacts, but documentation has another meaning: to collect, arrange, catalogue and retrieve all kinds of documents. By document is meant any communication made by man which has found a more or less permanent form. In all disciplines there is a continuous accumulation of various kinds of information, data, which at some stage will be used again to provide new knowledge, and these data must be arranged in such a way that they are easy to find again - information retrieval.

The entire archaeological material -artefacts and descriptions of investigated monuments­ consists of such data, communications of a more or less permanent nature. This material is already enormous and is growing faster and faster. Numerous principles of arrangement, of varying practicability, corresponding to the equally numerous points of view involved in subsequent treatment of this material, are in existence. None of them ensures that the desired material can be retrieved irrespective of the approach. The ideal is the registration and arrangement of the necessary information in such a way that it is easy to find, whichever aspect is to be studied.

An advanced and central documentation should seek to eliminate the following:

1) The eventual loss of information which is stored privately or purely mentally.

2) The inaccessibility of information which is recorded in an idiom which cannot be fully understood by an outsider.

3) The repetitive elementary cataloguing which is the beginning of much research.

4) The tedious or fruitless searching of different museum store-rooms for analogous material which has been dispersed according to the particular arrangement employed in storage.

5) Wear and tear on objects and archives caused by repeated searching.

Just as mechanization transferred much labour from a purely muscular to a controlling role, an advanced documentation should free scholars from the tedious and repetitive searches of large collections of material and give them time and energy for more creative work. But just as artisans were opposed to the mechanization of their trades and consequent devaluation of their skills, opposition can only be expected from many whose memory will be made partly superfluous by such documentation, which is basically a mechanization of memory.

The question is thus not whether there should be a National Index to carry out documentation in the future, but rather how this work should be organized and how it should function. To clarify this problem, it will be necessary to look at the data and methods of archaeology. In respect of the latter, it can be stated in general that the basis for the conclusions one may extract from the material remains, which are preserved in the form of artefacts or monuments, is the study of similarities and differences, what Sophus Müller called archaeological comparison [1].

There are many kinds of similarity. Similarities of external form and decoration are not the only ones: there are also affinities of material, mode of manufacture, size and context. Similarity cannot be measured directly and there are many degrees between dissimilarity and identity, but when groups are drawn up, the number of similarities can be counted. The smallest units which can exhibit similarities and differences are called by Malmer [2] typological elements, by Moberg [3] elements of similarity, but here merely elements. They can be divided into elements of shape, of proportion, of decoration of material, and of mode of manufacture [2].

The affinities we employ in studying archaeological material are 1) similarity between combinations of elements, expressed in types, and 2) similarity between combinations of types (artefacts and monuments) which are embodied in units of various kinds: cultures, periods, phases, horizons, etc.

A type can be defined by the number of well-defined elements which repeatedly occur together in a group of objects in a limited geographical area and inside a limited period. The number of such elements drawn into the definition can vary, and if it falls below a certain minimum, it would be natural to speak not of types, but of species.

The types we work with are usually defined from only a few elements, but this does not mean that a description can ignore the other elements. These may later form the basis for a subdivision of the type -chronological, geographical or functional- or for an entirely different typological division of the material when it is approached from another direction. It is thus impossible to decide beforehand which elements can be omitted in a description of archaeological material.

The concept of type has an essential function in the study of material, as it is with its help that one makes generalizations, whereby equivalent observations made in connection with several examples of the defined type are taken to be valid for all, for example dating, determination of the place of manufacture, determination of function -as long as there is nothing to invalidate them directly.

The task of creating a National Index for Danish artefacts and monuments, originally projected by Harald Andersen, Mogens Ørsnes and the author on entirely traditional lines, has now been taken up by the Institute of Prehistoric Archaeology and Ethnography of the University of Aarhus. The main register in this National Index will be a card index in format A4 (29. 7 X 21.0 cm), fig. 1, where each card will record information and carry a drawing and/or photograph of one artefact or one monument. This part of the project, the development of the traditional index cards, has been commenced, and simultaneously we are attempting more systematically to establish which needs a National Index should and can meet if it is to be at all practicable. In order to ensure that the assembled information will be clearly expressed and easily accessible -the prerequisite if a system is to have any permanence- it has been necessary to study the problems of registration and description affecting related disciplines, the various indexing systems, etc. Electronic data processing has also opened new perspectives for information retrieval.

An ordinary card index is merely a subject -divided list of documents which has been split up into cards with one document, for instance a book or an artefact, on each card.

One has thereby achieved an open system where new documents can be inserted and new subjects added. If this system is to work satisfactorily however, there are certain rules which have to be followed: the subject-matter has to be divided into a series of main categories, which may only be subdivided one at a time in a hierarchical system.

The subject-matter of the National Index is artefacts and monuments, each of which can be divided into a number of main categories, species, which are recorded in a register or list of species. As a sub-division of the species, for example axes, one would have the different types of axe. This requires, however, that true definitions of type are employed, and these have been virtually non-existent in archaeological literature until recently. Malmer's complaints of non-definition [4] are fully justified, and if one is to catalogue a large material with no other purpose than to create an order which will make it easy to find things, definitions are an absolute necessity.

The typological division of the Danish material is based on Sophus Müller's "Ordning af Danmarks Oldsager" 1888-95, which was superseded in respect of the Stone and Bronze Ages by "Danske Oldsager" 1948-53, edited by Therkel Mathiassen. These works provide illustrations of "all Danish artefact types" together with a short description, which, however, does not include any clear delimination of the types. In identifying an artefact, one will therefore often be in a position where certain affinities will call for allocation to one type and others to another type. The type divisions for one species are often based on quite different criteria, for instance core axe and flake axe (mode of manufacture) contra round-butted axe (shape) or silver-sheet fibula (material) contra relief-fibula (decoration) and fibula with semi-circular plate (shape).

The rule for the addition of new subjects, types, to a card index is that this may only occur by sub-division of one category at a time, in our case of one descriptive element at a time. One step in the classification of pots can thus comprise the neck: concave, straight or convex, and the next lugs: absent, one, two, three or more lugs. It will facilitate searching if the most frequently employed categories, the descriptive elements, are divided first. In the above example, pots with one lug must be sought in three different groups, if we let lug-classification be subordinate to neck-classification, whereas we would only have to look in one place, if neck-classification were subordinate. It is apparent that the more descriptive elements are employed, the more difficult it is to find a particular element in a hierarchical card index. One further example will make this clearer. In a telephone book, where the names are ordered alphabetically, we have a hierarchical arrangement of a large material. In the case of artefacts, we cannot know beforehand which category should be placed first, corresponding to the first letter in a name, as we may need to find any element, rather as if we had to find in the Danish telephone directory all the names ending in -sen. An index of artefacts or monuments built up on these hierarchical principles would become increasingly difficult to work with as it grew. The difficulties involved in the actual description are equally formidable.

The ordinary description of artefacts frequently makes use of a division into a series of naturally deliminated parts, which are then described separately: for example a pot can be broken down into body, shoulder, neck, rim, base and lugs. There is though, neither agreement as to how the individual parts should be demarcated nor as to the meaning of the terms used in description. On index cards photographs and drawings can to a certain extent replace a description, but often there are observations which cannot be illustrated. Illustration of all objects would also be an expensive and time-consuming process, but the main objection is that it does not contribute to cognition, which actually depends on a division into elements and a naming or classification of these. If we are to have an index within a reasonable period of time, which will fulfil the need for documentation, it will not be possible to draw or photograph everything, and we shall be forced to make do with descriptions which are as complete as possible.

The description system of Jean-Claude Gardin

The French archaeologist Jean-Claude Gardin [5], director of the Centre d'Analyse Documentaire pour l'Archéologie (CADA) under the Centre National de la Recherche Scientifique (CNRS), has been engaged for several years (see bibliography) on the same problems of documentation which we are now encountering, and has shown in several publications how these problems can be solved. Gardin starts from the principle that the most accurate and complete description comprises the definition and organization of a large number of elementary features, and to this end he has evolved a general procedure, an analytical description, employing a special language of symbols.

First Gardin lays down rules for the orientation of the object to be described, so that reference can be simplified: right/left, upward/downward, front/back, for example. Pots, for instance, are placed with the base horizontal and tools with the long axis horizontal, the functional part to the right and the handle to the left.

Next he divides the object in largely traditional manner into parts, for example neck, body, base, etc., in the case of pots. This segmentation is to a certain extent given by normal linguistic practice, which is itself an expression of the similarity of form occurring in each of the main functional categories. These categories, here called species [6], are for example, fibulae, pots, axes, ploughs. For the sake of clarity of description, however, it is necessary to lay down clear rules for the segmentation of each species into parts. Examples of such a segmentation are seen in figs. 7 and 12. Besides shape, ornament, material and mode of manufacture, there are other data which must be included in the description the nature of the find, i. e. grave, settlement, hoard, etc., and its context, i. e. whether it is an isolated find or whether it can in some way be combined with other objects by being of a single find, of a documented combination, or of a structure with a clear stratigraphy, i. e. together with, over or under other objects one would wish to feature in the description.

The last part of the system of description is the differentiation which consists of a break­down of the variations in which the different parts can occur. In Gardin's experience [7], one can generally be satisfied with broad divisions only providing a small number of qualitative differences. These are easily observed (e. g. straight, convex, concave -parallel, divergent, convergent- larger than, equal to, smaller than) and together cover the entire range of variation of each part. Quantitative elements -absolute or comparative values- need only be used to a minor extent and when they are employed, are arranged in groups of varying range according to the material being treated, for example, length can be grouped < (less than) 5 cm, 5-10 cm and > (more than) 10 cm.

The vocabulary necessary for a description of this kind is remarkably small, usually less than 40 terms used in each of the parts into which the object is divided. The parts and their various shapes are expressed by symbols, so that the description of each part comprises two or three symbols, for example Bc, of which the first capital refers to the part, here the body of a pot, and the small letter and/or number to one of the categories this is divided into, here with smooth transition from lower to upper half. The traditional synthetic terms have thus been replaced by analytical formulae.

These descriptive formulae are ideally suited to transference to punched cards and data processing machines. Gardin has also shown that they permit a simple solution to the problem of arranging the catalogue cards, as these can simply be numbered and arranged numerically under each species. Then an index is compiled to this numerical catalogue, containing one card for each of the variations into which the different parts are divided. This feature card gives the numbers of the requisite numbers in the catalogue. This system is called "inverted filing". On a special kind of index card -the so-called "peek-a-boo" cards- all the catalogue numbers are printed in advance. One denotes that a certain number in the catalogue has the property in question by punching the card in the number position, fig. 2. In this way, one can retrieve combinations of a whole series of properties.

The meaning of the symbols and the rules for segmentation are given in the code which is accompanied by an explanatory commentary.

The code is divided into chapters, denoted by capital letters, corresponding to the parts of the objects, and in each chapter there is a list of the possibilities for expressing the variations in each part. Two examples will give an impression of the nature of the code. The first is from the pottery code, chapter B (the body of the pot), fig. 3 illustrating the profile of the body. In most of the symbols employed there is a mnemotechnical content: v = concave, d = droit, x = convexe, i = divergent, u = parallele (the shape of the letter), o = convergent. The schematic vessel profiles in the figure are thus denoted by vi/xo, the profile being read from below upwards.

Fig. 4 refers to the conjunction between the upper and lower parts of the body: c = curved or continuous transition (continu), I = angular or discontinous transition (I is the normal mathematical symbol for an angle), q = discontinuous transition which combined with x (convex) to qx denotes a moulding and with v (concave) to qv a fluting, z = a broken line which combined with i (diverging) to zi denotes a diverging angle and with o (converging) to zo a converging angle. From the pottery code's chapter E (the neck), fig. 5 refers to the conjunction between body and neck and we can see that the meaning of the symbols is exactly as above. This is a good example of one of the principles of coding: the choice of a description so simple that it can be employed on different parts and also on different kinds of objects.

In other cases, as for example in chapter J (the cross-section of the lug or handle), the variations are merely numbered, fig. 6.

The actual analysis is not so difficult as the development of the code, but can naturally be rather a lengthy process if it covers a large or very complex material. The work is reduced with the aid of special forms for analysis, which contain directions and often a printed list of alternatives which have merely to be underlined, so that a single form can be completed in a few minutes or less. These systematic analyses take much less time than a normal description and are moreover both comprehensive and precise. To minimize the inevitable subjectivity involved in observation, the analysis is carried out by two persons independently and the result checked by a third. A sample pottery analysis form is shown in fig. 8.

The second of Gardin's artefact codes, the code tools and weapons, is constructed on the same principles. Here, the descriptive parts are divided into two groups, one comprising the functional parts and the other the hafting or handle parts, fig. 12, and each group is divided as usual into a series of chapters with capital letter headings, thus:

I General

A Functional part

B Hafting

C Dimensions, absolute and relative

II Functional part

D Section, faces

E Section, sides

F Outline, upper side

G Outline, lower side

H Outline, "end" and conjunction of sides and "end" I Details

III Hafting part

J Section

K Outline, upper side

L Outline, lower side

M The shape of the termination in a plane parallel to the functional part

N The shape of the termination in a plane at right-angles to the functional part

0 Appendages of the hafting part.

P Details

IV Connections between II and III Q Conjunction of K and F

R Conjunction of L and G

S Conjunction of N and E

The terms "upper" and "lower" are defined according to the rules for orientation mentioned above.

Two examples from this code will show that it is not very different from the pottery code and that several symbols are used again with the same meaning.

The first example, fig. 9, is from chapter A and shows a division into functional types, merely numbered in succession. Fig. 10 is from chapter D (cross-section of the functional part). Note that several of the symbols have been employed before in the pottery code with the same meaning: d = straight, c = curved or continuous transition, 1 = angular or discontinuous transition, n = discontinuous transition with angles in direct succession, Ø = ending in a point. Fig. 11 is from chapter F (the contour of the top of the functional part). To make it possible to describe a change in the curve, the contour is divided into two parts, each with its own symbol in the vertical column where the terms d = straight, x = convex, and v = concave are seen in different combinations. In addition, the course of the curve may be characterized by three terms Ø, I and II, denoting that the end points of the curve and the point where it changes course are in the same plane, Ø, or that the latter point ties respectively above, I, or below, II, the line joining the two end points. Each of these denotations is then augmented in three columns, according to whether the point where the curve changes direction lies to the left (<), in the middle (=) or to the right (>).

Also the implement code has its special forms for analysis, divided as for the pottery code into parts (capital letters) and a number of variables (small letters and numerals). A sample completed description sheet for an implement is reproduced in fig. 13.

*

The development of systems of description like those mentioned here can be a valuable contribution to archaeological scholarship by establishing a firm conceptual system, which should be the basis of the national systems, which are under constant revision. Gardin's analytical system is not dependent on the language in which it is compiled, as the conceptual content of the symbols is well defined and constant, irrespective of the language employing them. This is by no means true of normal archaeological concepts. For instance, the term microliths comprises in 1) Danske Oldsager I, no. 3, trimmed implements manufactured from blades 2-5 cm long and about 0.5 cm wide, though nos. 76-91 of the same work place the width between 0.5 and 1 cm; 2) according to Malmer [10], trimmed implements manufactured from blades with a maximum length of 5 cm and a width less than, or equal to, 1/5 of the lenght, i. e. 1 cm; 3) according to Childe [11], small implements less than 1 (or 1.5) inch (2.5-3.8 cm) long, which may have been manufactured from large blades -the definition must also include transverse points; and 4) according to Therkel Mathiassen [12], besides what Danske Oldsager calls microliths, also broad trapezes and rhomboids (oblique points) and to a certain extent transverse points.

This example is not unique. It is, so to speak, in the nature of the concepts employed that they are always subject to reformulation in the light of new material and new points of view. This must be so, as is the case in other disciplines, for example zoology, botany and geology, but all these have regular and objective systems of description, a language of description, which does not depend on the approach to the material. It is in this language that the most advanced concepts are expressed, and it serves as an excellent subject­index when material with particular characteristics is to be found.

Archaeology has not yet developed such a language, with analytical concepts instead of the usual synthetic ones, but Jean-Claude Gardin's system can be the beginning.

The need for an analytical language of description has existed for many years. It can be informative to look at one attempt in that direction, the one made to create an international system of description for ploughing implements at the 1954 Conference in Copenhagen [13]. One wanted to find forms of description which were independent of the existing plough typology (p. 37, Bratanić; p. 161, Aitken). A. Steensberg suggested (p. 44) that ploughs should be divided into descriptive parts, each denoted by a capital letter; this would be followed by a number referring to one of the categories or types to which the part had been allocated by a committee of experts. Bratanić (p. 48), who had inspired Steensberg to make this suggestion, wanted in addition to the analytical system a shorter system employing synthetic nomenclature.

The object of this system would be (p. 46) to ensure easy access to comparative material, to assist description by enumerating all parts and variations, to give clear definitions of these, and to ensure that the collected data were described in a uniform manner which could therefore be understood by others.

The Centre d'Analyse Documentaire pour I'Archéologie so far has developed only two codes of artefact description, namely those for pottery and metal implements. At the moment no systematic registration is planned similar to that projected in Denmark. The aim of the work in France has first and foremost been to develop universally applicable principles of description and it has therefore been applied to different cultural fields -iconography (oriental seals, Greek vase paintings, Roman mosaics), texts (the Koran, Akkadian tablets)- and has developed codes for these. The Centre is also occupied with a description of religions and secular buildings. It has been demonstrated that the principles outlined above will probably be applicable to practically all cultural material.

Development of codes

As will be apparent from the description of the various parts of the system -catalogue, index, code, commentary- the codes are a prerequisite not only for the verbal description, but also for the arrangement of these descriptions. Before systematic registration can commence, it will thus be necessary to develop a whole series of codes, and we ought therefore to look at some of the problems involved.

As mentioned above, the method employed in developing the codes comprises a reduction of the variety of the objects by breaking these down into well-defined pa4rts and establishing the variations occurring in each part. These elementary designations are relatively few, but may, however, occur in innummerable combinations.

To assess the value of such a system, several criteria should be considered. From a scientific point of view, descriptive terms should both represent a complete objectivity and contain the possibility of formulating by combination a general or detailed description of any object. From a logical point of view, an analytical system should strive towards a simple and uniform description and towards one and only one description of each document (artefact or structure), which should also be as brief as possible.

It is clear that the segmentation into parts and the differentiation of these parts determine the practicability of the system, and that such division can only be attempted after a close study of a representative croos-section of the material. The greater part of the material which the Danish National Index will be concerned with has for many years been the subject of scientific study, which will be of great value in the development of the codes. A result of this study has been more or less precise segmentation into parts, for instance a sword's blade, hilt, upper and lower guards, pommel, etc. There are, of course, many possibilities in segmentation. Pot profiles could, for instance, be divided into a series of equal portions [14] which would be described separately, but by this means the profile is not broken down into simple shapes and one of the criteria for deciding whether a division is good or bad is just the amount of simplification of description which results.

Existing studies of material should also be given due consideration when the variations in each part are to be coded. These variations can be considered a continuum which, in spite of its smooth transitions, is nevertheless in normal linguistic practice broken down into groups. This is the case, for instance, with the lines p. 115, which in normal usage are divided into convex, straight, and concave, without the boundaries between these designations being clear. How straight a line must be in order to be called straight may for example be expressed by the largest permitted deviation from the straight line between two points, in per cent of the length of the line.

There are thus many possible ways of dividing this row of lines, but none of them will be entirely appropriate under all conditions, i. e. afford the simplest description or definition of any type. Every type -definition is dependent on the elements, the language, available for expression. To obtain types comprising as many elements as possible, it can be useful to reduce or extend parts of the variation range of the same elements, as shown in fig. 14, where a type can be defined by the elements c and f, but be made more comprehensive by qualification with the elements (a-), (d-), (e-), and (g-). By this means the division of the variation range for these elements has become dependent on the type, and the divisions no longer have a constant value. Clearly, the more the limits of the variation range depend on a few types, the less universal the system of description, the code, will be.

The problem of designing a code which is not only universal but capable of expressing the whole variety of types, can be solved, at least theoretically, by making the elements smaller, so that what we now call elements are expressed as compounded of a few simple units with many combination possibilities. Such a procedure would enlarge the code considerably and in consequence the time necessary for analysis, as one can no longer be satisfied by easily observable qualitative differences, but must to a large extent employ quantitative methods requiring measuring instruments. Only practical experience will show whether this elaboration of the system is necessary.

The division of variations should as far as possible be hierarchical, so that when the index is to be used, one can make a fine or a coarse sorting, just as one can in the topographical index [15] where counties, hundreds and parishes represent different levels of description. This topographical code is also "artificial", and bears no reference to the actual distribution of archaeological material. It is clear that a division into units as large as counties will not be adequate to show the distribution of any archaeological type. By sub-dividing twice, here into hundreds and parishes, the units become small enough for any distribution to be described. Thanks to this hierarchical structure, one can quickly find out which of the analysed objects have the characteristic of having been found on the island of Funen, for instance, as it is only necessary to examine the two index cards of the two counties on Funen -Odense and Svendborg- instead of cards for 15 hundreds or 202 parishes.

In the same manner, a code can be developed for gold bracteates, for example. The first step is the division of the ornament into men, animals and other devices; in the second division men are divided up into head, body, arms and legs; the third step divides the human head into hair, eyes, nose and chin; in the forth step the hair is divided into different styles, 1, 2, 3, .....

The number of elements which can be brought into the description are for any object so many, that they can be considered infinite. New details and combinations of details are constantly isolated for further study, and in registration one cannot hope to enumerate everything which future scholarship will be interested in. That the registered material will nevertheless be of use is due to two important features of the system:

1) The analysis comprises the object as a whole and not merely a few selected portions of it and

2) the code has been developed to cover the whole variation range of each part. The whole object is thus mapped out so that with an ordinary knowledge of the different species of objects and their codes, one can reach certain conclusions about details which to a greater or lesser extent will be connected to those groups into which the code is divided. If one had described, for example, a collection of pottery containing collared flasks according to Gardin's pottery code, which has been developed in ignorance of the existence of these flasks, one would know anyhow that these vessels would have to be sought in the group which has a thickening of the neck.

If one wished to investigate the occurrence of an element which could not, unlike the previous example, be directly connected to information in the standard analysis, the National Index could still be of use. This is because it would contain a complete list of the material in which the element in question could be present, and would permit a representative selection of this material to be extracted, on which it could be investigated whether the presence of the new element could possibly contribute anything towards the solution of the problems centering on this material. If this investigation was negative, the results could nevertheless be stored for future use.

By the degree of differentiation of the variation groups one can decide how detailed the analysis is to be. A highly detailed analysis would not only result in each object being represented by one and only one description, but would also reach the point where each description could be applied to one and only one object! By this means one would be in the position of describing differences instead of similarities. To bring out the similarities, it can be advantageous to group together variations of elementary features instead of attempting to separate the whole range of variation, so that, for example, more or less concave profiles were merely described as concave.

The disadvantage of this procedure is that a search of the index would yield a list of objects which were not exact parallels to what one was looking for, but merely had a certain resemblance to it. This is, however, not entirely a disadvantage, as it leaves to the individual researcher how much of the retrieved material to use.

Special features are noted in the description only to the extent that they can, alone or in combination with other features, contribute to a separation of certain groups of objects which experts consider to form a single entity or be of special interest. For example, in a pottery code for Danish material it would be advantageous to include a feature such as the collar on collared flasks.

Automatic classification

A National Index constructed on the basis of analyses of the kind described here would not only be a comprehensive guide to the archaeological material, but could also be the starting point for an automatic classification, i. e. a division of the material into similarity groups with the aid of electronic computers. Classification is one of the basic tasks of any science. Observations and objects must be arranged in same way or other before the reason for their similarities can be discovered and used as a basis for prediction. Many phenomena occur in such numbers and in so many variations that it is unlikely that they can be understood without being systematized.

The material remains -artefacts and monuments- are an example of such phenomena, and the types (cf. p. 123) a form of classification of them. It is true that, as stated above, the basis for the separation of types is their similarities, but these are often selected in the light of a particular interpretation of the material and are thus not objective, as certain characteristics are given more weight than others. The development of types is not a classification of the total material, but only of a selected, characteristic part. This classification employs, moreover, only a few elements, a so-called monothetical classification, primarily because the human brain can only comprehend the variation of a few elements at a time.

Electronic computers have opened up new horizons, in that we here have an instrument capable of processing large numbers of variables according to a previously planned procedure -the so-called program. The development of programs for this kind of classification, so-called polythetical classification, has created new interest around the classification problem, for what makes one classification better than another, and in criteria for "natural classification".

With the aid of electronic computers different groups of similarities can be isolated by means of different programs. This procedure has been applied to archaeological material on several occasions [16] and the Centre de Calcul de la Maison des Sciences de l'Homme, Paris, has experimented with several different classification methods.

As a foundation for such an automatic classification, a description matrix is developed, where the horizontal rows represent the objects, whilst the variations in the parts are registered by the vertical columns. In such a matrix, the objects can be arranged so that those showing the closest resemblance feature together, and the elements can be arranged so that the similarities of different objects occur as close to one another as possible. A rearranged matrix of this kind thus features series of elements and series of objects, which can both represent a chronological order.

The fact that automatic classification results in a grouping of the material at some variance with the types archaeologists have worked with hitherto has been responsible for its rejection by many scholars, but the fact remains that the very foundation of archaeological research is the study of similarities, and an arrangement of the material according to similarities in a so-called taxonomic system must be a basic archaeological operation.

The National Index

It is the intention to employ the principle of analytical description developed by Jean­Claude Gardin in the Danish National Index. This will thus comprise:

1. The catalogue -numerically arranged cards (fig. 1) with drawings and/or photographs, information about locality and conditions of the find, and about storage locality.

2. The index to the catalogue. This will in the first instance comprise cards of the peek-a-boo type [18] (fig. 2).

3. The code containing an organized enumeration and definition of the terms used in the index.

4. The commentary to the code, explaining the principles and mode of use of the latter.

It is clear that a whole series of codes is necessary if the complete archaeological material of artefacts and monuments is to be described. Each code normally comprises the group we have called a species (p. 124), but can also comprise several species. It would be helpful to have a system of description which also gave a clear differentiation of the material into species, since these have never been defined, although one is not normally in doubt as to the limits.

In order to obtain a standard arrangement of the cards according to species, and to connect the terms in normal usage with the codes, a register or list of species is developed, where the species are collected in coarse groups according to function, e. g. jewelry, means of transport, weapons. To the register is appended an alphabetical list of terms for species and types used in the literature, with a reference to the corresponding terms used in the register. The species are provided with symbols which can be used on the catalogue cards in listing the objects included in find combinations (p. 125).

As stated, the object of the National Index is to effect a registration of the archaelogical material from the entire country and provide a subject-index which will make this material more easily accessible to research.

The registration and systems of description are intended to aid future research and cannot await further work on the material, but must rather assist in this work. The systems of description cannot therefore be final, but must be elaborated or simplified in accordance with the level of scientific investigation.

As the development of the codes, the systems of description, not only demands a profound knowledge of previous treatment of the material, but also of the current attitudes, it will be appropriate to develop the codes in co-operation with those specialists who want a particular material registered and described with a view to further research. In this way, the widest consideration could be given to those particular features which the specialist wishes to be included in the code. After the completion of the scientific investigation, suitable changes could be made in the code, which would thereby at any time express the actual stage of research. The specialist will often be interested in a smaller portion of the material than that which we have called a species (p. 124) and it can therefore be necessary to work with several specialists. As it is of great importance that the code can also be used on foreign material, it will also be necessary to co-operate with foreign experts.

After the development of the code, the actual work of registration and description will be carried out by non-scientific personnel. In all cases where an object cannot be described directly according to the code, it will be left to a specialist to decide whether to alter the code or give rules for how the object can be fitted to the existing code.

Besides the description, drawings will be made of the whole material or of selected objects, depending on the nature of the material. This will be done by special draughtsmen. The punched-card index will also be maintained by non-scientific personnel. The staff of the National Index can thus be built up of groups each consisting of 1 qualified archaeologist, 1-2 assistants, 1 draughtsman and 1 clerk.

Apart from the systematic work of registration and analysis, it is also the intention to register all photographed and drawn (measured) artefacts and monuments. These illustrations are primarily to be found in the archaeological literature, but the quantity stored away in museum archives is by no means small.

Establishment of the National Index will not only affect the scientific work but also certain parts of the ordinary work at museums. Closely bound up with the National Index are thus the storage arrangement and accession catalogues.

The objects in museum store-rooms can be arranged according to many different principles, but it is an almost general rule, at least in Denmark, that objects from a closed grave find, settlement find, hoard, or other single entity are kept together. The general arrangement is made by period, locality or merely by the order in which the objects are received by the museum. If the collection is very large, a combination of principles may be applied.

In order that the National Index may live up to its pretensions and be an effective register of the archaeological material, it is necessary that references can be obtained from the index not only to the catalogue (the numerically arranged cards), but also to the actual objects, which must therefore have an unambiguous storage classification in the various museums and collections. The easiest way of achieving this is by arranging material in store-rooms according to one or more of the main entries on the card -nature of the find, period, locality, or inventory number (museum number)- and by adopting the same definitions for these units in the museums and the National Index.

The object of an accession catalogue is primarily to secure the connection between the numbered object and the information one has about its find context. To ensure that the object can be identified if the number should disappear, the accession catalogue contains a very brief description of the object, plus any peculiarities which may serve to identify it. Apart from the purely metrical information and the designation of kind, the general description is just as variable and inexact as all other artefact descriptions, and it is worth considering its replacement by an analytical description of the type described above. The accession catalogue could then, besides the special identification description (which might be replaced by a photograph) merely comprise a designation of species and a reference to the object's number in the National Index catalogue for that species. In this connection, electronic data processing could usefully be employed, as is shown in a plan worked out by Nordiska Museet in Stockholm [21).

The primary objective of the National Index is to create an effective register of the artefacts and monuments which form the basis of prehistoric research, thus ensuring that all the existing material becomes accessible to this research.

Apart from this important documentational aim, the National Index holds other possibilities, of which the employment of the codes as the basis of a special archaeological language and of automatic classification, has been briefly mentioned.

The National Index brings us into direct contact with a fundamental problem -the problem of description, which has received little attention in archaeology, where the study of artefacts and monuments has exclusively concerned itself with reconstructing changing aspects of the past: evolutional, ecological, sociological, etc.

To consider artefacts and monuments as a self-sufficient object of study, is a point of view corresponding to the reorientation of philology, structural linguistics, which L. Hjelmslev formulated in 1943. Hjelmslev's conception of the linguistic sign as an "entity, generated by the connexion between an expression and a content" (p. 47) can be applied to material remains, which can also be considered as signs the expression of which we can describe and classify, but the content of which we can only attempt to understand by using analogies. Hjelmslev pointed out that it is not only the content which is worth studying, but that a study of the expression, the linguistic sign, could lead to a knowledge of the nature of these signs, and he himself remarked (p. 107 seg.) that other sign systems, for instance folk-costume, could be studied in the same way. Thus the material remains can also be considered a sign system, which one should attempt to grasp not as a conglomerate of non-material phenomena (e. g. geographical, phsychological, sociological), but as "a self-sufficient totality, a structure sui generis" (p. 5 seq.). The way to this cognition is to provide "a procedural method by means of which objects of a premised nature can be described self-consistently and exhaustively" (p. 15).

Olfert Voss