Theoretical and Methodological Considerations in Central European Neolithic Archaeology Proceedings of the ‘Theory and Method in Archaeology of the Neolithic (7th - 3rd millennium BC)’ conference held in Mikulov, Czech Republic, 26th – 28th of October 2010 Edited by Jan KoláU František Trampota BAR International Series 2325 2012 Published by Archaeopress Publishers of British Archaeological Reports Gordon House 276 Banbury Road Oxford OX2 7ED England bar@archaeopress.com www.archaeopress.com BAR S2325 Theoretical and Methodological Considerations in Central European Neolithic Archaeology. Proceedings of the ‘Theory and Method in Archaeology of the Neolithic (7th ‐ 3rd millennium BC)’ conference held in Mikulov, Czech Republic, 26th – 28th of October 2010 © Archaeopress and the individual authors 2012 ISBN 978 1 4073 0908 8 Copy‐editing by Robert Brukner The preparation of this book was co‐funded by the Institute of Archaeology and Museology, Masaryk University, Brno, Czech Republic Printed in England by 4edge Ltd, Hockley All BAR titles are available from: Hadrian Books Ltd 122 Banbury Road Oxford OX2 7BP England www.hadrianbooks.co.uk The current BAR catalogue with details of all titles in print, prices and means of payment is available free from Hadrian Books or may be downloaded from www.archaeopress.com gravE typology and Chronology of a lEngyEl CulturE sEttlEMEnt: forMalizEd MEthods in arChaEologiCal data proCEssing Peter Demján Department of Archaeology, Comenius University Gondova 2 SK-81499 Bratislava Slovakia peter.demjan@gmail.com Abstract: The application of typological methods is a well-known means to ind regularities within an archaeological assemblage. A formalized approach is needed when dealing with larger sets of data to eliminate bias arising from deining types a priori and searching for structures which we already expect to exist. By creating a quantitative descriptive system and using multivariate statistic methods we are able to identify similarities and dissimilarities between single artefacts as well as between individual ind complexes. These similarities (or factors) represent chronological and cultural relationships as well as post-deposition transformations. We can further sort ind complexes using seriation (e.g. correspondence analysis) to elaborate a chronological sequence. This paper presents an application of a formalized typological and chronological method on graves and their inventories from the Lengyel Culture settlement in Svodín (Southwest Slovakia) using multivariate statistical analysis. Keywords: Pottery analysis, typology, relative chronology, multivariate analysis, Lengyel Culture, settlement burials. important of these sites - Svodín (Southwest Slovakia) provided the basis for this paper. Excavated between 1971 and 1983 under V. Němejcová-Pavúková, it yielded inds from 111 graves, two rondels and several hundred settlement features dating to stage I of the Lengyel culture. While the material from rondels has already been published (Němejcová-Pavúková 1995), the settlement features and graves await publication. In this paper the methods and techniques used for processing the rich burial inds within the settlement will be presented. Introduction In the area of western Hungary and southwest Slovakia a new cultural phenomenon appears around 4800 BC -the late Neolithic Lengyel cultural complex. It is characterized by a distinct pottery, rich in shape repertoire, and with a broad palette of painted motifs. The relatively large and densely populated settlements consist of three-part longhouses, often with a central circular ditched earthwork - the so-called “rondel”. Inhabitation spreads to new natural environments, occupying higher elevation sites and the distance between individual settlements increases. Development within social structures is also relected in changing burial practices. Settlement burials which appeared during the preceding middle Neolithic Želiezovce culture are now the only form of burial we ind. Lengyel sites provide us with valuable evidence about the development in both ritual and everyday life of Late Neolithic society. Unfortunately, only a handful of settlements with burials from the irst stage of this culture have been excavated and published to date (Dombay 1960; Kalicz 1985; Němejcová-Pavúková 1986a; NěmejcováPavúková 1995; Neugebauer-Maresch et al., 2002).We expect new additions when material from recent rescue excavations in Hungary is published. Approaching burials within settlements When processing inds from settlement burials we have to keep in mind speciic elements which differentiate these features from those graves found in a regular burial ground located outside of a settlement. The irst question to pose is whether we are really dealing with burials inside the occupied area of the community, or rather graves or grave groups located in vicinity of the settlement, placed eventually on abandoned habitation. Considering that a settlement can grow or shrink over time, or it can be abandoned and later resettled, we must take into account the possibility that a nearby burial ground could be absorbed by its boundaries, or new burials can be situated in a previously inhabited area. If the archaeological evidence is lacking reliable stratigraphic relationships between the graves and settlement structures, the question of whether The scarcity of information about early Lengyel settlements with burials makes it all the more pressing to process the hitherto unpublished excavations and make data available to the scientiic community. The inds from one of the most 77 Grave Typology and Chronology of a Lengyel Culture Settlement we are dealing with true settlement burials or with a burial ground situated nearby the settlement is not trivial. extract information about the past living culture contained within this assemblage, we need to correctly identify the transformation processes which led to their present shape and location. The second issue concerns the chronological assignment of graves and their inventories. We cannot assume a linear or radial trend in burying the dead as would be the case on a regular burial ground. Therefore graves from earlier and later phases of the settlement can be found literary side by side. In order to determine a chronological sequence we need to make use of dating methods which do not rely on stratigraphic or spatial observations. Transformation processes Transformation processes can essentially be divided into three categories: pre-deposition, deposition and postdeposition (Neustupný 2007, 51-54). In the case of the Svodín burials we are dealing with fairly well preserved material from closed contexts (i.e. graves), which means post-deposition processes did not substantially inluence the composition of the (non-organic) inventories. Therefore the deposition transformation is the one we must consider further in our analysis of the assemblage. It represents the process of transition of the artefact from the living to the dead culture. The reason for this transition is very important - for grave inventories it is an intentional exclusion from the world of living and deposition into the ground with ritual motivation. The selection of particular grave goods may therefore not necessarily be a representative sample of the material culture of the community which deposited them. A distinct predominance of certain pottery types in graves (pedestaled bowl, mushroom vessel) when compared to assemblage from settlement pits may mirror this selection process. Application of typological method for determining a detailed chronology Assigning artefacts to distinct groups based on their visual or technological similarity - the typological method - is a well-established approach to evaluating archaeological inds. This technique can lead to the elaboration of a typological sequence relecting the chronological order in which the types appeared. A traditional approach to typology is to irst determine main categories (or classes) of inds by empirical observation of the material. We can then proceed to differentiate distinct types and variants within these categories by inding similarities and dissimilarities between the single artefacts. A different approach is to consider the limitations of the entirely empirical approach, which is to a great degree inluenced by observer bias, and elaborate a typology based on a statistical correlation between different attributes, by which we can describe the inds (Hodder and Hutson 2003, 182). An attribute is a property of an artefact, which can be meaningfully enumerated (either by quantifying, measuring or indicating its presence or absence). We must keep in mind that this process still involves a certain degree of subjectivity, especially in the selection of the attributes to be measured and choice of measurement techniques (Neustupný 2007, 122-123). The nature of post-deposition transformations of burial inds is quite different than that of the inds from features like pits or ditches. Material in features which have been continuously used over a longer period of time is subject to complex repositioning and mixing due to cultural and natural inluences. In contrast to that grave goods often remain undisturbed after their deposition, subject only to natural decay processes and occasional secondary intrusions. In this aspect the process of excavation and documentation of the inds also has to be considered as the de facto inal post-deposition transformation. In our effort to achieve a chronological ordering of the burials from Svodín we must therefore keep in mind the possibility that distinct types of grave goods as well as the burial rite itself were subject not only to temporal changes but also to other hitherto unknown factors. We also have to consider that the changes of different attributes of the grave inventory over time happened at different frequencies. It is possible that some aspects of the burial rite were approached more conservatively then other and their simultaneous presence in two closed contexts does not necessarily mean their contemporaneity (cf. Hodder and Hutson 2003, 177). After elaborating the typological description of our material, we can proceed to identify chronological relationships between types. The techniques for this are based on the assumption that the common occurrence of artefacts in a closed ind context (e.g. a grave or a hoard) also implies contemporaneity. Under ideal circumstances (i.e. if simultaneously deposited artefacts really represent a particular ‘style’ which was commonly deposited in graves only for a limited time span) the sequence resulting from a seriation of the assemblage will correspond with the chronological order of appearance of the types and the deposition of the artefacts. Even after taking all aforementioned factors into consideration, the information about the past living culture that we are able to extract will refer more to the diachronic development of the burial rite than the actual chronological phases of the settlement. To elaborate a complete intrasite chronology, we would have to include also all other Such ideal conditions rarely if ever occur in real ind assemblages. The archaeological inds we are evaluating represent a dead culture - an assemblage of artefacts existing in the present, representing a living culture of the past but separated from the world in which they were created and used (Neustupný 2007, 47). To be able to 78 Peter Demján settlement structures, such as pits and ditches, into our analysis. Zalai-Gaál applied seriation to chronologically sort burial inds from several Lengyel Culture sites (Zalai-Gaál 2007). In the latter study a formalized method was also used to elaborate a typology based on quantitative attributes of ceramic inds. Monothetic vs. polythetic structures The presence of non-chronological relationships in a ind assemblage is a fact of which archaeologists are well aware. There are different ways to approach this multidimensionality of data but they all have in common an effort to reduce the number of dimensions (i.e. commonalities appearing across the ind complexes) whilst trying to highlight the cultural and chronological aspect and suppress the aspects determined by the different transformation processes. An example for this is the elaboration of a typology of copper axes based on their shapes and dimensions while ignoring their colour which is a product of patination. Chronological ordering using seriation depends largely on the selection of chronologically sensitive attributes of the artefacts. The selection process can be considered empirical and is often based on stratigraphic observations. This approach can reduce the seriation to a conirmatory technique used only to validate the chronological relations which we already identiied empirically. If on the other hand we attempt to include all potentially chronologically sensitive attributes in our seriation (e.g. the decoration of pottery, the shapes of different parts of the vessels) we risk that other than temporal dimensions of the assemblage (like the inluence of gender or other social factors on the grave inventory) will also affect the resulting order and will be misinterpreted as a chronological development. The search for hidden structures in archaeological data is described by E. Neustupný (2007, 128-130) as a synthesis of structures, whereas he distinguishes between monothetic and polythetic structures citing among others the groundbreaking work of D. Clarke (1968). A monothetic structure is characterized by one particular attribute or group of attributes the presence of which is necessary for an artefact to be considered a member of this structure. An example of a monothetic structure is the linking of grave inventories containing boar tusks with the male gender. The aforementioned pitfalls when searching for hidden structures can be avoided by using multivariate methods. One of them is factor analysis (FA) - a statistical technique used to ind important commonalities (factors) in the assemblage similar to the more widely used principal component analysis but putting higher emphasis to commonalities (Shennan 1997, 303-305; Neustupný 2007, 137). The factors extracted by this method represent polythetic structures and using a regression method we can calculate scores representing the pertinence of each unit (ind or ind complex) to a particular structure. Structures (or dimensions) identiied in the assemblage in this way can then be explored separately and chronological ordering achieved without the risk of mixing attributes representing different structures (e.g. attributes of grave inventory connected with social status and attributes which represent a change of decoration style over time). The usage of a hierarchical monothetic approach can be considered an evolutionary step in deining archaeological structures. Based on a speciic attribute (e.g. the presence of a pedestal on ceramic vessels) the inds are classiied into categories, which are then further divided into subcategories based on other monothetic criteria (e.g. an s-shaped neck proile). A polythetic structure is deined by the probability of the presence of certain attributes. Its main difference from a monothetic structure is that an artefact does not necessarily need to have all the deining attributes present to be considered a member of a structure. The traditional empirical typological method approaches the polythetic model by allowing exceptions when deining types. By judging the relationships based on individual attributes the similarity and dissimilarity of artefacts can be assessed and polythetic structures deined. Multivariate analysis as opposed to a simple application of seriation can be described as an explorative technique resulting in the synthesis of new structures which represent chronological or other relationships between the units of the assemblage. Based on the multidimensional approach where each dimension represents a particular attribute of an artefact, E. Neustupný proposed the name ‘vector synthesis’ for this family of methods to better relect its means and goals (Neustupný 2007, 137). We could consider every unit of the explored assemblage represented by a row of the descriptive matrix to be a vector in a multidimensional space the coordinates of which are the metric or quantitative attributes of the artefact or ind complex. Using multivariate techniques (e.g. factor analysis, multiple correspondence analysis or cluster analysis) we synthesize the information contained in the descriptive matrix and try to ind new vectors which best represent the structures in a lowerdimensional space. Formalized typological and chronological method One of the irst formalized methods applied in archaeology was seriation. When used correctly the resulting ordering of rows and columns of the descriptive system will relect a chronological ordering of the ind complexes and their attributes (for a more comprehensive description, see e.g. Orton 1980). The work of E. Kazdová (1984) with material from Tešetice-Kyjovice (in southern Moravia) is an example of the application of this technique in processing an Early Lengyel ind assemblage. From more recent studies, I. An important example for application of multivariate 79 Grave Typology and Chronology of a Lengyel Culture Settlement Figure 1. Archaeological Data Manager software used for data entry, ind measurement and classiication. statistic techniques to elaborate a formalized typology and chronology is the work of J. Macháček (2001; 2010) on early medieval ceramic material from Pohansko (southern Moravia). The paper of J. John concerning the structure of the grave inventory of the Linear Pottery Culture (John 2005) as well as other papers inspired by the seminal work of E. Neustupný published in the book Příspěvky k archeologii 2 (Neustupný and John 2005) should be mentioned. 1. Search for hidden structures based on similarities and dissimilarities within the ind assemblage and ind complexes. 2. Elaboration of a typological classiication of the inds and ind complexes based on the identiied structures. 3. Ordering of the individual types and ind complexes in a chronological sequence. Before selecting the optimal techniques for processing the material from Svodín, output requirements had to be considered. First, there is a need to catalogue inds because the assemblage from the burials in Svodín have not yet been published in their entirety. The descriptive system must contain all the necessary data to produce a catalogue with textual and pictorial information about the individual burials and their inventories. The next requirement is a classiication of the inds using a formalized method, which is also the irst step required to identify structures within the whole ind assemblage representing the relicts of the past living culture. These two requirements imply also the next step: the elaboration of an internal chronology of the development of the burial rite on the settlement. The irst step in creating a descriptive system was the digitization of all available data. Drawings and photographs of the inds were digitized into raster images and text was transcribed into electronic form. Excavation plans were digitized and a geodatabase was created based on them using ArcGIS. Descriptive system of the ind assemblage In consideration of the need to quantify data and create queries based on different input requirements of statistical software packages a relational database was chosen as the platform for creating the descriptive system. MS: Access allows easy data entry and manipulation, is compatible with ArcGIS, and it is possible to create queries using SQL which is often necessary for more complex quantiication of nominal data. The analysis of the ind assemblage unfolds in three steps: 80 Peter Demján Field Name Description ID Unique identiier feature_nr Feature number trench Trench number side Body placement side (left or right) orientation Orientation (azimuth) sex Sex of the deceased age Age of the deceased (in coded form) X coordinate of the grave centre (WGS 1984 UTM 34N) Y coordinate of the grave centre (WGS 1984 UTM 34N) coords_x coords_y Nominal descriptors for non-ceramic inds are limited to a description of the ind type in coded form. For ceramic inds, the shape of the plastic elements (knobs) and the distribution and motif of painted decoration on the vessel is described in a coded form. Metric descriptors represent the size of the inds and, if applicable, their quantity. For ceramic vessels, measurements of different parts of the body are taken and stored in these descriptors. These attributes are necessary for the elaboration of a typology using multivariate statistical techniques. The deinition of which parts of the vessels should be measured was adapted from the Lengyel pottery typology Numerical code of Moravian painted ware (Podborský et al., 1977, 31-34) by modifying it so that it best relects the perceived visual image of the vessel. This is of course a point where the author’s subjective choice is involved in the process of elaborating typology. The ceramic vessels were irst divided into seven basic body types based on which metric attributes can be measured (Figure 4). The conical or convex shape was expressed as a measure of convexity (-1 meaning a concave, 0 a conical or cylindrical and +1 a convex shape). A degree of proile inclination was calculated where applicable using the formula (diameter1 - diameter2) x 5 / height suggested by V. F. Gening (1977, 97). In the descriptive system uniied ields are used for the attributes of all vessel body types according to the ield descriptions in Figure 3. The points of measurement for different vessel body types and formulas for calculating proile inclinations are deined as follows: Figure 2. Selected descriptors from the table of graves. Streamlining of the data entry, veriication, processing and publishing process was a major issue in this work. It was necessary to process a relatively large amount of data in a short time span and only one person was available for all tasks. Since there was no specialized software for archaeological data management available at that time which would suit all the above requirements a simple Archaeological Data Manager (ADM) software was developed (Figure 1). It works as an interface to anMS Access database, enhancing its functionality by streamlining the typological classiication and measuring of inds. It also offers the ability to graphically display the inds or ind complexes in the database and localise individual inds within ind complexes. Textual and pictorial data from the descriptive system can then be compiled into a catalogue in PDF format ready for publication. The possibility of displaying inds as a list sorted and iltered by any variable enables us to visually inspect the results of the ordination using different statistical methods. This serves as useful feedback and makes it easier to spot errors in data entry. 1. One-part vessel • Rim diameter • Neck height: Vertical distance between the rim and the base. • Base diameter • Neck proile inclination: (rim diameter - base diameter) x 5 / neck height 2. Pedestaled one-part vessel • Rim diameter • Neck height: Vertical distance between the rim and the base. • Base diameter: Diameter of the base of the upper part (not pedestal base). • Pedestal base diameter • Pedestal height: Vertical distance between the base of the vesel and the pedestal base. • Neck proile inclination: (rim diameter - base diameter) x 5 / neck height • Pedestal proile inclination: (pedestal base diameter - base diameter) x 5 / pedestal height 3. Two-part vessel • Rim diameter • Neck height: Vertical distance between the rim and the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck. • Carination diameter: Diameter on the point where the convexity of the body changes or the concavity The inds database consists of two main tables: the table of graves (Figure 2) and the table of grave inventory (Figure 3). Relations between these tables and other ancillary tables used for the coding of variables are based on unique identiiers (ID). The description of grave inventory is accomplished using a single table containing both ceramic and non-ceramic inds. Some of the descriptors are relevant for both types of inds and some only for ceramic vessels. They can be divided into three basic categories: localization, metric and nominal descriptors. The table also has some ancillary ields necessary to form relations with different data tables and for cataloguing (feature number, ind number etc.). Localization descriptors carry the information about the placement of the ind inside the grave. Besides Cartesian coordinates there is also a nominal description of the placement in relation to the body of the deceased which is more useful when searching for structures in the placement of grave goods. 81 Grave Typology and Chronology of a Lengyel Culture Settlement Field Name Description ID Unique identiier feature_nr Feature number ind_nr Find number body_type Body type of ceramic vessel (in coded form, see Figure 4) Localization descriptors placement Placement relative to body (in coded form) placement_detail Placement detail (in coded form) placement_x X coord. of the ind in the grave (in cm) placement_y Y coord. of the ind in the grave (in cm) Metric descriptors height Find height (according to orientation on drawing) (in mm) width Find width (according to orientation on drawing) (in mm) amount Number of unique pieces (e.g. beads of a necklace) units Number of units (e.g. blades) Nominal descriptor ind_type Find type (in coded form) Metric descriptors relevant for ceramic inds rim_diameter Rim diameter (in mm) neck_diameter Neck diameter (in mm) neck_height Neck height (in mm) belly_diameter Belly diameter (for body types 4, 6, 7) or carination diameter (types 3, 5) (in mm) upper_body_height Upper body height (in mm) lower_body_height Lower body height (in mm) base_diameter Base diameter (for body types 1, 2, 3, 5, 6) or carination diameter (types 4, 7) (in mm) pedestal_base_diameter Petestal base diameter (for body types 2, 5) or base diameter (types 4, 7) (in mm) pedestal_height Pedestal height (for body types 2, 5) or base height (types 4, 7) (in mm) neck_shape Neck convexity (-1 to 1) neck_carination Carination below the neck (1 or 0) upper_body_shape Upper body convexity (-1 to 1) upper_body_carination Carination on the belly (1 or 0) lower_body_shape Lower body convexity (-1 to 1) pedestal_shape Convexity of the walls of the pedestal (for body types 2, 5) or the base (types 4, 7) (-1 to 1) neck_proile Neck proile inclination upper_body_proile Upper body proile inclination lower_body_proile Lower body proile inclination pedestal_proile Proile inclination of the walls of the pedestal (for body types 2, 5) or the base (types 4, 7) Nominal descriptors relevant for ceramic inds painted_type Distribution of the painted or incised decoration on the vessel (in coded form) painted_upper_motif Motif of the decoration on the upper body (in coded form) painted_lower_motif Motif of the decoration on the lower body (in coded form) painted_inside_motif Motif of the decoration on the inside (in coded form) plastic_type Type of plastic elements (in coded form) Figure 3. Selected descriptors from the table of grave inventory. 82 Peter Demján • • of the neck begins or there is a carination below the neck. • Lower body height: Vertical distance between the belly or carination and the base. • Base diameter • Neck proile inclination: (rim diameter - belly diameter) x 5 / neck height • Lower body proile inclination: (belly diameter base diameter) x 5 / lower body height 4. Two-part mushroom vessel • Rim diameter • Neck height: Vertical distance between the rim and the maximum body diameter. • Belly diameter: Maximum body diameter. • Lower body height: Vertical distance between the belly and the carination below the belly. • Carination diameter: Diameter on the carination below the belly. • Base height: Vertical distance between the carination below the belly and the base. • Base diameter • Neck proile inclination: (rim diameter - belly diameter) x 5 / neck height • Lower body proile inclination: (belly diameter carination diameter) x 5 / base height • Base proile inclination: (base diameter - carination diameter) x 5 / base height 5. Pedestaled two-part vessel • Rim diameter • Neck height: Vertical distance between the rim and the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck. • Carination diameter: Diameter on the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck. • Lower body height: Vertical distance between the carination and the base. • Base diameter: Diameter of the base of the upper part (not pedestal base). • Pedestal height: Vertical distance between the base of the vessel and the pedestal base. • Pedestal base diameter • Neck proile inclination: (rim diameter - carination diameter) x 5 / neck height • Lower body proile inclination: (carination diameter - base diameter) x 5 / lower body height • Pedestal proile inclination: (pedestal base diameter - base diameter) x 5 / pedestal height 6. Three-part vessel • Rim diameter • Neck diameter: Diameter of the narrowest part between the rim and the belly or the carination below the neck. • Neck height: Vertical distance between the rim and the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck. Belly diameter: Maximum body diameter. Upper body height: Vertical distance between the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck and the belly. • Lower body height: Vertical distance between the belly and the base. • Base diameter • Neck proile inclination: (rim diameter - neck diameter) x 5 / neck height • Upper body proile inclination: (belly diameter neck diameter) x 5 / upper body height • Lower body proile inclination: (belly diameter base diameter) x 5 / lower body height 7. Three-part mushroom vessel • Rim diameter • Neck diameter: Diameter of the narrowest part between the rim and the belly or the carination below the neck. • Neck height: Vertical distance between the rim and the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck. • Belly diameter: Maximum body diameter. • Upper body height: Vertical distance between the point where the convexity of the body changes or the concavity of the neck begins or there is a carination below the neck and the belly. • Lower body height: Vertical distance between the belly and the carination below the belly. • Carination diameter: Diameter on the carination below the belly. • Base height: Vertical distance between the carination below the belly and the base. • Base diameter • Neck proile inclination: (rim diameter - neck diameter) x 5 / neck height • Upper body proile inclination: (belly diameter neck diameter) x 5 / upper body height • Lower body proile inclination: (belly diameter carination diameter) x 5 / base height • Base proile inclination: (base diameter - carination diameter) x 5 / base height Typological classiication of pottery shapes The irst step in typological classiication of ceramic inds is to divide vessels into basic classes according to different metric attributes. This step is necessary so that the individual vessels are comparable by those attributes. In the case of the material from Svodín seven basic classes have been identiied (Figure 4). For each basic class of vessels which contained a suficient amount of specimens typological structures were identiied using factor analysis. This was done using SPSS Statistics. The descriptors used were all metric attributes relevant for the particular basic vessel class. A detailed description of performing a factor analysis of archaeological assemblages 83 Grave Typology and Chronology of a Lengyel Culture Settlement Figure 4. Classiication of ceramic vessel body types according to measurable attributes. 84 Peter Demján Descriptor belly_ diameter rim_diameter neck_height lower_body_ height lower_body_ shape base_diameter lower_body_ proile neck_ carination Factor 1 0,9321 0,9080 Descriptor lower_body_ proile neck_shape 0,8677 0,8172 neck_proile neck_ carination 0,2744 0,2375 neck_height 0,2134 0,1912 rim_diameter -0,0174 0,1222 belly_diameter -0,1407 0,0985 base_diameter -0,2767 0,8506 0,6765 neck_shape 0,0143 neck_proile -0,5922 lower_body_ shape lower_body_ height technique attempts to sort the specimens into clusters based on their calculated distance in a Euclidean space where the selected descriptors are represented as coordinates. The process of clustering creates clusters of the closest (and thus most similar) specimens irst, then groups those clusters into larger clusters and so on, until the whole assemblage can be sorted in a hierarchical tree structure (represented by a dendrogram). For more detailed information on this method see the work of C. Orton (1980, 47). Ward’s method is particularly well suited for clustering of continuous numerical data like the factor scores which were produced in the irst step of our pottery analysis (Shennan 1997, 241-245). The settings used in SPSS Statistics were Ward’s method, and measuring of interval as squared Euclidean distance. Finally a dendrogram plot was produced. The descriptors used in this step were the factor scores extracted in the previous step except factor 1 which always represented a sorting of the vessels by size. Based on the dendrogram produced by clustering subtypes and variants were deined. For example for vessels of body type 3 two classes, four subclasses and eight variants were identiied (Figure 7). The irst digit of the type code represents the body type of the vessel, the second and third the class and subclass and the fourth the variant. Factor 2 0,2653 -0,3963 -0,6487 Figure 5. Loadings on factors 1 and 2 extracted for vessels of body type 3 (two-part vessel). is given in the works of E. Neustupný (1993; 2007) and S. Shennan (1997). The settings used were: calculation of a correlation matrix using coeficients, extraction of factors using principal components method based on eigenvalue greater than one, Varimax rotation and calculation of factor scores using regression. In the case where less than three factors were extracted based on the eigenvalue, the ixed number of factors to extract was set to three. Otherwise typological sorting would be dificult as the irst factor always represents a sorting of the vessels by size and is therefore discarded from further steps when elaborating a typology. The other factors extracted in this step represent commonalities in the assemblage according to which the ceramic inds can be further divided into classes, subclasses and variants. As an example we can see two factors extracted from metric attributes of vessels of body type 3 (two-part vessels) (Figure 5). Factor 1 represents a similarity of the vessels based on carination diameter (here the variable belly diameter), rim diameter, neck height and lower body height. The negative loading for neck proile inclination relects the fact that vessels with wider and higher upper parts tend to have a less inclined neck proile. Factor 2 shows a connection between the lower body proile inclination and the convexity of the neck. The negative loading for lower body height indicates a less probable occurrence of vessels with high, strongly inclined lower body and a convex or conical neck. It is a good practice to visually inspect the results of factor analysis by sorting the inds by their factor scores (Figure 6). This way any errors in data entry can be identiied and we can also verify our interpretation of the factors. By examining the factor scores of the vessels in different typological groups, typical factors and descriptors for those groups were identiied and described (as is shown in Figure 8). We have to keep in mind however that such a description it is just a simpliied interpretation of the structures and cannot be considered an explicit deinition of a ceramic type as is common when using a monothetic approach to typology. The factors and the scores calculated for each vessel in the descriptive system represent only a measure of probability with which the specimen belongs to a particular structure. The clustering into distinct ceramic classes and their subsequent description was done largely to compare the presence of structures identiied from metric attributes of the pottery, with other structures which will be identiied in the descriptive system of the burial contexts based on their nominal attributes. It is also impossible to directly compare factors extracted for vessels of different body types, because they represent different descriptors depending on which parts of the body were measured. Lastly, a classiication of the inds together with verbal description helps us to compare our material to that from other sites where a monothetic typological approach was used. Typological classiication of burial contexts In a manner similar to identifying structures in pottery shapes using factor analysis and their subsequent typological classiication we can search for hidden structures in complexes represented by graves and their inventories. The attributes of non-ceramic grave goods are dificult to describe by metric values. When dealing with plastic elements (knobs) and painted decoration an empirical approach to typology appears more appropriate Further classiication of the pottery into classes, subclasses and variants was conducted by applying hierarchical cluster analysis (HCA) using Ward’s method. This multivariate 85 Grave Typology and Chronology of a Lengyel Culture Settlement Figure 6. Vessels of body type 3 (two-part vessel), sorted by factor scores on factor 2. High positive values are typical for vessels with strong inclination of the lower body proile and a conical neck. Negative values are typical for shapes with higher lower body. than trying to ind a formalized method. The small number of specimens of each kind of non-ceramic inventory makes a more detailed classiication unnecessary. Ceramic variants which occurred less than three times were joined with other variants of the same subclass where possible. The quantiication of nominal attributes describing the position of grave goods relative to the body required further generalization of some variables as distinct types of grave goods were only rarely positioned on the exact same position in different graves. The reduction in number of variables was achieved by two steps. First, by reducing the ceramic classes to six basic shapes (one-part bowl, two-part bowl, pedestaled bowl, mushroom vessel, pot and beaker), next by joining the 52 identiied positions into 18 larger clusters. These generalizations were possible also due to a quite regular habit of the placement of the deceased in lexed position on the right side. Quantiication To be able to analyse graves from Svodín using multivariate methods we irst need to quantify the nominal attributes of their inventories. The attributes in this case are classes of the ceramic vessels, painted decoration motifs, types of plastic elements and types of non-ceramic grave goods. All these nominal attributes can be expressed numerically by quantifying them, i.e. counting the number of specimens of each type in each grave. The rows of the resulting descriptive matrix will represent the individual graves and the columns will represent different types of decoration, ceramic classes etc. The applicability of this method on archaeological inds was demonstrated by L. Chroustovský (2008). Classiication Factor analysis was then applied on the descriptive matrix using SPSS Statistics with settings identical to those used for pottery analysis. It was performed separately for descriptors of ceramic classes, decoration motifs, plastic knob types, non-ceramic types and positions of grave goods. First only two factors were extracted and extreme outliers identiied. These atypical ind complexes were The descriptive matrix was created in MS Access by creating a query for the table of grave inventory using SQL. Only attributes occurring three and more times in the assemblage were quantiied to keep a statistical relevance. 86 Peter Demján Figure 7. Typological classiication of vessels of body type 3 (two-part vessel) according to factors 2 and 3 using Hierarchical Cluster Analysis. Class hierarchy Code Typical factors class 3100 -2 subclass 3110 -2, -3 variant 3113 -3 then no longer used for the extraction of factors from the particular type of descriptor. The reason for the elimination of extreme outliers is that they constitute categories of their own and suppress the more subtle structures present in the whole assemblage. Luckily only one grave was an outlier in all ive groups of descriptors and therefore could not be used for the overall analysis of the burials. However, the information contained in the shapes of its rich ceramic inventory contributed to the elaboration of pottery shape typology. After eliminating the outliers, factors with eigenvalues greater than one were extracted for each of the ive descriptor groups and factor scores were calculated for each grave. Factor description high lower body no carination, high lower body no carination variant 3114 -2 high lower body subclass 3120 3 carination below the neck variant 3121 -2, 3 carination below the neck, high lower body, wide base variant 3122 -2 high lower body variant 3123 3 carination below the neck class 3200 2 convex neck, strong lower body proile inclination subclass 3210 2 convex neck, strong lower body proile inclination variant 3211 2 convex neck, strong lower body proile inclination subclass 3220 -3 no carination variant 3221 -3 no carination Based on the factor scores a classiication of the graves using hierarchical cluster analysis was performed (with settings the same as when classifying pottery) and using the dendrograms graves were assigned to clusters according to different types of descriptors (e.g. a grave could be in cluster 6 according to the painted decoration of its inventory and in cluster 3 according to types of plastic knobs). Figure 9 shows an example of clustering of graves according to plastic elements on the ceramic grave goods. These clusters can be described by their typical factors and descriptors similarly to the description of pottery classes and variants Figure 8. Ceramic classes, variants and subvariants of vessels of body type 3 (two-part vessel). 87 Grave Typology and Chronology of a Lengyel Culture Settlement Figure 9. Typological classiication of graves according to factors 1-6 describing the occurrence of distinct types of plastic elements on the ceramic inventory using Hierarchical Cluster Analysis. 88 Peter Demján Cluster Typical factors represent the membership of a grave in a particular group. MCA was performed using the PAST software package created by Ø. Hammer, D. A. T. Harper and P. D. Ryan (2001). The result is a scatter plot where the irst and second axes represent the irst two dimensions of the reduced multidimensional space and the dots represent individual graves and grave types (Figure 11). By mapping the sex of the deceased on the scatter plot it becomes apparent that the second axis sorts the grave inventories by gender. We can therefore hypothesize that the irst dimension represents the chronological order. While a deinitive validation of this hypothesis would require the presence of stratigraphic relations between the graves (which is not the case in Svodín) we can attempt at least a partial validation of the chronological sequence by using external evidence. This is also important to determine the orientation of the time axis i.e. which end represents the earlier and which the later burials. Factor description small - indistinct and anthropomorphic knobs 1 2 2 3, 2 vertically oblong and round knobs with horizontal opening 3 4 saddle-form knobs with vertical or horizontal ridge 4 1 atypical (less than 3 graves) non-speciic (no typical factors for this cluster) 5 6 6 horizontally oblong knobs 7 5 drawn-out knobs Figure 10. Typological grave groups (clusters) deined according to the occurrence of distinct types of plastic elements on the ceramic inventory. It is reasonable to think that burials did not take place in regular intervals one after another. We must therefore assume that there are luctuations in our ordering based on highly fragmented information. To be able to assess the relative-chronological assignment of the individual graves and their types we must irst identify gaps in the MCA ordering on the horizontal axis. The clusters of graves between these gaps represent typological-chronological phases, in which there is a higher probability that the member graves are roughly contemporaneous (Figure 12). We can now identify structures which are typical for these phases and compare them with inds from other early Lengyel sites. with the same limitations concerning such a monothetic interpretation of polythetic structures (Figure 10). The presence or absence of an attribute typical for a group in a grave does not automatically assign it to that particular group. Determining the chronological sequence The method of relative chronology is based on the assumption that certain cultural phenomena are subject to diachronic change. We already identiied relics of these structures in the assemblage from Svodín using explorative statistical techniques. If we order these ind complexes in such a way that those belonging to the same structures (represented by typological grave groups) will be close to each other, the resulting sequence should represent their relative chronological ordering. External evidence The Lengyel phase of the settlement in Svodín encompasses only one ceramic stage (Lengyel I), which limits our search to structures which appear only for a brief time span. By deinition such structures will be present in relatively small numbers in the ind assemblages which further complicates their identiication. It must be noted that despite the small number of excavated settlements from the Lengyel I stage the state of processing and publication of the material is quite good. Detailed internal chronologies have been elaborated for the ceramic inds from Tešetice and Kamegg using formalized methods (Kazdová 1984; Doneus 2001). The ceramic grave inventories from Zengővárkony, Friebritz and several other sites were evaluated in a summarizing study, in which a formalized typology of the pottery was elaborated based on its metric attributes and a seriation of the ind complexes was performed (Zalai-Gaál 2007). The search for chronological links to Svodín was based on these works. Also of great use were the extensive works of J. Pavúk on Lengyel pottery typology (for the latest see Pavúk 2007; 2009) and the typological analysis of the Svodín material by V. Němejcová-Pavúková (1986a; 1986b; 1995). The burial inds from inside the rondel in Friebritz (Neugebauer-Maresch et al., 2002) also provided A multivariate statistic method well suited for this application is multiple correspondence analysis (MCA). Similarly to cluster analysis, it uses the Euclidean distance of data points in multidimensional space to highlight similarities and dissimilarities within a descriptive system. Its advantages lie in its ability to work with nominal variables (which in our case represent the membership in a typological group) and the plotting of the resulting ordering in the form of a two-dimensional scatter plot (Shennan 1997, 308). Various examples of the application of this method in archaeology can be seen in P. Pavúk (2010). A good example of the elaboration of a chronological sequence using MCA can be seen in the work of J. Macháček (2001; 2010). The rows of the descriptive matrix which we have to prepare to use this method represent the individual graves and columns represent typological groups (by non-ceramic ind types, ind placement, painted decoration, plastic elements and ceramic classes). The values 1 and 0 in the matrix 89 Grave Typology and Chronology of a Lengyel Culture Settlement Figure 11. Sorting of the graves (black dots) according to their classiication to distinct typological groups (grey dots; by nonceramic ind types, ind placement, painted decoration, plastic elements and ceramic classes) using multiple correspondence analysis (MCA). Figure 12. Identifying four typological-chronological phases of the settlement burials based on their chronological relations represented by the horizontal axis of the MCA scatter plot. 90 Peter Demján important information about the absolute dating of this stage of the Lengyel Culture. Feature nr non_ cer place paint plast body phase MCA order 3 8 2 5 2 1 1 2773 3 10 2 5 2 1 2 13480 2 5 6 6 2 1 3 17181 2 2 2 6 2 1 4 12280 3 8 6 3 2 1 5 13280 6 5 5 5 2 1 6 10680 2 5 5 5 2 1 7 4574 2 5 5 6 5 1 8 672 2 5 2 6 5 1 9 4876 2 2 2 5 2 1 10 7678 2 2 2 5 3 1 11 3774 2 2 2 2 2 1 12 171 2 5 2 2 5 1 13 2973 2 5 5 3 2 1 14 15 10580 Some dificulty lies in the fact that the material analyses in the aforementioned work rely on a more or less monothetic typological approach. Chronological assignment is directly connected with the appearance of speciic attributes on the pottery to which it is dificult to ind parallels in the polythetic structures identiied in Svodín. The focus of this article on methodical procedures does not provide space for elaborating the several different parallels and analogies which ultimately led to the orientation of the time axis and a validation of the chronological sequence. Analogies in ceramic vessel shapes and most importantly in chronologically very sensitive motifs of painted decoration put the typological-chronological phases I and II of the Svodín burials into the early phase (IA) of the Lengyel Culture. At least for phase I we can assume a contemporaneity with the Friebritz settlement (ixed by an analogous polythetic structure in painted decoration, found on both sites), which is dated into the phase Lengyel IA not only by relative-chronological observation but also by 14C dating (Stadler and Ruttkay 2007, Figure 5). Based on the analysis of the grave inventory it cannot be ruled out that the settlement lasted only during the phase IA of the Lengyel Culture, the duration of which according to radiocarbon measurements is c. 113 years (Stadler and Ruttkay 2007, Table 7). The resulting chronological sorting of the graves and typological groups can be seen on Figure 13. Examining the spatial distribution of graves from different chronological phases over the area of the settlement shows that there is no linear or radial trend in the burials and all observed grave clusters contain graves from all phases. This implies that we are dealing with burials which were an integral part of the living settlement. Further analysis of the material from other settlement features in Svodín and stratigraphica relationships of graves and pits, houses and rondel ditches, may shed more light on the relation between the living community and its dead. Conclusions Choosing a formalized approach to typological and chronological analysis of a large assemblage has proven to have some clear advantages over the traditional empirical approach. It is possible to eliminate observer bias in many steps of the analysis especially when it comes to pottery classiication. The ability to work with true polythetic structures instead of just an advanced monothetic model enables us to identify even subtle structures within the assemblage and helps us distinguish between chronological and other cultural or technical aspects. Application of a combination of multivariate statistic methods allows us to compare attributes with very different frequencies of occurrence in the assemblage and where quantiication is possible also to include both metric and nominal attributes in our analysis. A well designed 2474 2 2 2 7 5 1 12180 2 7 6 3 5 2 16 10380 2 5 2 3 2 2 17 14481 2 5 2 3 2 2 18 9479 3 3 3 3 2 2 19 12780 2 5 5 4 1 2 20 17882 8 2 2 3 1 2 21 17281 2 2 2 5 1 2 22 17682 9 5 5 7 1 2 23 4174 2 2 2 6 4 2 24 271 2 5 2 7 2 2 25 1673 2 5 3 2 2 2 26 972 2 5 3 7 2 2 27 2173 2 5 3 7 1 2 28 3974 2 6 3 5 5 3 29 9379 5 7 3 5 1 3 30 3073 2 7 3 5 4 3 31 15981 5 7 3 5 2 3 32 17782 2 5 3 5 4 3 33 7778 2 6 3 5 1 3 34 7078 5 1 3 5 2 3 35 3574 2 5 2 5 4 3 36 14081 7 5 1 5 4 3 37 11280 1 1 1 1 1 4 38 13981 7 6 4 3 2 4 39 2373 2 6 2 5 2 4 40 16081 2 6 2 5 2 4 41 13781 2 6 2 5 2 4 42 11380 5 6 4 1 4 4 43 16381 2 5 4 1 2 4 44 13080 4 4 4 4 3 4 45 n Group typical for phase I n Group typical for phase II n Group typical for phase III n Group typical for phase IV Figure 13. Chronological sorting of the graves and their membership of different typological groups. 91 Grave Typology and Chronology of a Lengyel Culture Settlement formalized process of typology makes it possible to enter new data into the descriptive system even by non-experts and at a relatively fast pace. Here the use of metric attributes to describe vessel shapes presents a great advantage. John, J. 2005. Příspěvek ke struktuře pohřební výbavy v kultuře s volutovou keramikou (A contribution to the structure of LBK Culture grave goods), in E. Neustupný and J. John (eds.), Příspěvky k archeologii 2, 11-19. Plzeň, Aleš Čeněk. Kalicz, N. 1985. Kőkori falu Aszódon. Múzeumi Füzetek 32. Aszód, Petői Múzeum. Kazdová, E. 1984. Těšetice-Kyjovice I. Starší stupeň kultury s moravskou malovanou keramikou. Brno, Univerzita J. E. Purkyně. Macháček, J. 2001. Studie k velkomoravské keramice. Metody, analýzy a syntézy, modely. Brno, Masarykova univerzita. Macháček, J. 2010. Zur Methode der Bearbeitung der (frühmittelalterlichen) Keramik aus Siedlungsarealen, in B. Horejs, R. Jung and P. Pavúk (eds.), Analysing Pottery. Processing - Classiication - Publication, 41-71. Bratislava, Universita Komenského v Bratislave. Němejcová-Pavúková, V. 1986a. Vorbericht über die Ergebnisse der systematischen Grabung in Svodín in den Jahren 1971-1983. Slovenská archeológia 34 (1), 133-176. Němejcová-Pavúková, V. 1986b. Charakter der NachLužianky-Entwicklung der Lengyel-Kultur. Béri Balogh Ádám Múzeum Évkönyve 13, 225-232. Němejcová-Pavúková, V. 1995. Svodín. Zwei Kreisgrabenanlagen der Lengyel-Kultur. Studia Archeologica et Mediaevalia 2. Bratislava, Universita Komenského v Bratislave. Neugebauer-Maresch Ch., Neugebauer J.-W., Großschmidt K., Randl U. and Seemann R. 2002. Die Gräbergruppe vom Beginn der Bemaltkeramik im Zentrum der Kreisgrabenanlage Friebritz-Süd, Niederösterreich. Preistoria Alpina 37, 187–253. Neustupný, E. 1993. Archaeological method. Cambridge, Cambridge University Press. Neustupný, E. 2007. Metoda archeologie. Plzeň, Aleš Čeněk. Neustupný, E. and John, J. (eds.) 2005. Příspěvky k archeologii 2. Plzeň, Aleš Čeněk. Orton, C. 1980. Mathematics in Archaeology. London, Collins. Pavúk, J. 2007. Zur Frage der Entstehung und Verbreitung der Lengyel-Kultur, in J. K. Kozłowski and P. Raczky (eds.), The Lengyel, Polgár and related Cultures in the Middle/Late Neolithic in Central Europe, 11-28. Kraków, Polska Akademia Umiejętności Kraków and Institute of Archaeological Sciences Eötvös Loránd University, Budapest. Pavúk, J. 2009. Die Entwicklung der Želiezovce-Gruppe und die Entstehung der Lengyel-Kultur. In A. ZeebLanz (ed.), Krisen – Kulturwandel – Kontinuitäten. Zum Ende der Bandkeramik in Mitteleuropa. Beiträge der internationalen Tagung in Herxheim bei Landau (Pfalz) vom 14-17. 06. 2007. Internationale Archäologie, Band 10, 249-266. Rahden, Verlag Marie Leidorf GmbH. Pavúk, P. 2010. Pottery Processing at Troy. Typology, Stratigraphy and Correspondence Analysis: How Do They Work Together, in B. Horejs, R. Jung and P. Pavúk After elaborating a chronological ordering in this way it is possible to retrospectively identify attributes typical for particular chronological phases. This information can be then used to date inds where a polythetic structure cannot be found due to lack of data. The fact that we derive monothetic descriptors from hidden polythetic structures revealed by the analysis and not vice versa underlines the exploratory nature of this process. Of course it has to be mentioned that this study relied on data from fairly well preserved and documented closed contexts and ceramic vessels which could be completely restored. If we were dealing with highly fragmented material from settlement pits and ditches, which is also subject to many post-deposition transformations, the elaboration of a reliable formal system would be more dificult. References Chroustovský, L. 2008. Kontingenční tabulky a vektorová syntéza : příspěvek k možnostem transformace nominálních dat a jejich následné matematické syntézy (Contingency tables and vector synthesis. A contribution on the potentional of nominal data transformation and its subsequent mathematical analysis). In J. Macháček (ed.), Počítačová podpora v archeologii 2, 49-60. Brno-PrahaPlzeň, Masarykova Univerzita, Archeologický ústav Akademie věd České republiky v Praze, Západočeská univerzita. Clarke, D. 1968. Analytical Archaeology. London, Methuen. Dombay, J. 1960. Die Siedlung und das Gräberfeld in Zengővárkony. Beiträge zur Kultur des Aeneolithikums in Ungarn. Archaeologia Hungarica 37. Budapest, Hungarian Academy of Sciences. Doneus, M. 2001. Die Keramik der mittelneolithischen Kreisgrabenanlage von Kamegg, Niederösterreich. Ein Beitrag zur Chronologie der Stufe MOG I der Lengyel Kultur. Mitteilungen der prähistorischen Kommission der Österreichischen Akademie der Wissenschaften, Band 46. Wien, Verlag der Österreichischen Akademie der Wissenschaften. Gening, V. F. 1977. Program statistického zpracování keramiky. In J. Malina, Nové metody popisu tvaru sbírkových předmětů. Morfometrika artefaktu, 91–97. České Budějovice, Jihočeské muzeum. Hammer, Ø., Harper, D. A. T., and Ryan, P. D. 2001. PAST: Paleontological Statistics Software Package for Education and Data Analysis. Palaeontologia Electronica [online] 4 (1), [Accessed 28 January 2011], 1-9. Available from: <http://palaeo-electronica. org/2001_1/past/issue1_01.htm>. Hodder, I. and Hutson, S. 2003. Reading the Past: Current Approaches to Interpretation in Archaeology. 3rd ed. Cambridge, Cambridge University Press. 92 Peter Demján (eds.), Analysing Pottery. Processing - Classiication Publication, 73-98. Bratislava, Univerzita Komenského. Podborský, V., Kazdová, E., Koštuřík, P. and Weber, Z. 1977. Numerický kód moravské malované keramiky. Problémy deskripce v archeologii. Brno, Univerzita J. E. Purkyně. Shennan, S. J. 1997. Quantifying Archaeology. 2nd ed. Edinburgh, Edinburgh University Press. Stadler, P. and Ruttkay, E. 2007. Absolute Chronology of the Moravian-Eastern-Austrian group (MOG) of the painted pottery (Lengyel-Culture) based on new radiocarbon dates from Austria, in J. K. Kozłowski and P. Raczky (eds.), The Lengyel, Polgár and related Cultures in the Middle/Late Neolithic in Central Europe, 117146. Kraków, Polska Akademia Umiejętności Kraków and Institute of Archaeological Sciences Eötvös Loránd University, Budapest. Zalai-Gaál, I. 2007. Zengővárkony-Svodín-Friebritz: zu den chronologischen Beziehungen zwischen den territorialen Gruppen der Lengyel-Kultur aufgrund der Gräberfeldanalyse, in J. K. Kozłowski and P. Raczky (eds.), The Lengyel, Polgár and related Cultures in the Middle/Late Neolithic in Central Europe, 147-184. Kraków, Polska Akademia Umiejętności Kraków and Institute of Archaeological Sciences Eötvös Loránd University, Budapest. 93