**Non-Linear Spatial Patterning in Cultural Site Formation Processes - The Evidence from Micro-Artefacts in Cores from a Neolithic Tell Site in Greece**

Dimitris Kontogiorgos

[49] Tsai, C. F., Tsai, C. W., Wu, H. C., & Yang, T. (2004). ACODF: A Novel Data Cluster‐ ing Approach for Data Mining in Large Databases. *Journal of Systems and Software*, 73,

[50] Lee, S., Kim, G., & Kim, S. (2011). Self-Adaptive and Dynamic Clustering for Online

[51] Alahakoon, D., , S., Halganmuge, K., & Srinivasan, B. (2000). Dynamic self-organiz‐ ing maps with controlled growth for knowledge discovery. *IEEE Transactions on Neu‐*

[52] Merkl, Rauber D., & Dittenbach, M. (2002). The growing hierarchical self-organizing map: exploratory analysis of high-dimensional data. *IEEE Transactions on Neural Net‐*

[53] Qin, A.-K., & Suganthan, P.-N. (2004). Robust growing neural gas algorithm with ap‐

[54] , L., Robert, K., & Warwick, K. (2002). The plastic self organising map. Hawaii. *Pro‐ ceedings of the 2002 International Joint Conference on Neural Networks, IEEE*, 727-732.

[55] Hung, C., & Wermter, S. (2003). A dynamic adaptive self-organising hybrid model for text clustering. Melbourne. *Proceedings of the Third IEEE International Conference on*

[56] Liu, Yuanchao, Wang, Xiaolong, & Wu, Chong. (2008, January). ConSOM: A concep‐ tional self-organizing map model for text clustering. *Neurocomputing*, 71(4-6),

[57] Liu, Yuan-chao, Wu, Chong, & Liu, Ming. (2011, August). Research of fast SOM clus‐ tering for text information. *Expert Systems with Applications*, 38(8), 9325-9333.

[58] Liu, Yuanchao, Wang, Xiaolong, & Liu, Ming. (2009). V-SOM: A Text Clustering Method based on Dynamic SOM Model. *Journal of Computational Information Systems*,

plication in cluster analysis. *Neural Networks*, 17(8-9), 1135-1148.

Anomaly Detection. *Expert Systems with Applications*, 38, 14891-14898.

133-145.

857-862.

5(1), 141-145.

*ral Networks*, 11(3), 601-614.

220 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

*works*, 13(6), 1331-1341.

*Data Mining, IEEE*, Florida, USA, 75-82.

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/51193

## **1. Introduction**

Micro-artefacts (i.e., cultural particles smaller than 2mm in diameter), due to their abun‐ dance and incorporation into the sedimentary matrix of an archaeological deposit, constitute a significant part of the cultural particles present [21]. Micro-artefact analysis is extensively complex due to the different micro-artefact categories that may appear in an archaeological context and also because of the numerous cultural (and non-cultural/natural) formation processes that may have been involved in the creation of characteristics specific to an ar‐ chaeological context.

Recently, the use of a non-linear method (i.e., spherical-SOFM) on micro-artefact data has shown that the method is able to recognize and to provide a visual representation of microartefact patterns prior to performing any statistical analysis on the data, providing a quick view into possible relationships or differences that may occur between temporally, spatially, and culturally different archaeological contexts (i.e., pits and ditches from the Neolithic Tell site at Paliambela (Pieria region-Northern Greece) which unusually comprises an extended settlement component [8].

It was shown that the spherical-SOFM non-linear method revealed patterns among the data that linear methods were unable to classify. Furthermore, the method attempted to over‐ come the difficulties posed by the friable nature of different micro-artefact classes (for exam‐ ple, unburnt clay, burnt clay, bone, shell, or charcoal). Material characteristics and the process of micro-artefact generation, including the effects of post-depositional processes, were considered as important factors in the search for strong pattern recognition [9]. The analysis has shown that similar classes of micro-artefacts in three analyzed data sets were

characterized by different non-linear associations, further suggesting that these were possi‐ bly formed through different cultural formation processes [8].

Greece) while section 4 and section 5 offer the results of this study and some concluding re‐

Non-Linear Spatial Patterning in Cultural Site Formation Processes

http://dx.doi.org/10.5772/51193

223

The Spherical Self-Organizing Feature Map (S-SOFM), introduced by Kohonen [7], maps ndimensional data into a low-dimensional space. The spherical SOFM [17] the low-dimen‐ sional space is a tessellated sphere that is formed by subdividing an icosahedron. Every vertex on this sphere is a strategic location of an n-dimensional vector that represents an en‐ semble of similar data vectors which are assigned to the vector during the mapping opera‐ tion. It is therefore necessary to visually enhance variations in the data using the physical attributes of the mapping lattice. The benefit of a spherical lattice in the implementation of the S-SOFM is that the enclosed space can be used to generate a 3D visual representation of

Conventional implementation of the S-SOFM method have used a 2D lattice as the low-dimen‐ sional space, and associations in the data are visualized by means of a terrain map, wherein el‐ evation represents some aspect of the vector(s) at that location [27, 24]. Relative similarity between data vectors mapped into the sphere can be visualized by introducing distortions in the sphere accompanied by changes in the colour. Informative characteristics of the data are re‐ flected as distortions and colour gradations on the surface of the sphere. The formulation of these measures is a non-trivial task and often application dependent. The measures reflect de‐ sired data correlations (either linear or non-linear) and must be defined by the researcher who is familiar with the underlying data set. It is this aspect of the S-SOFM that differs from existing literature about the self-organizing feature map. The S-SOFM utilizes the spherical lattice of the S-SOFM space to generate a visual form of the clustered data that is more intuitive and easy to perceive. A visual form of the data is created by scaling the radial distance of the vertices on the sphere in proportion to a measure characterizing some physical aspect of the data. Exam‐ ples illustrating the various implementations of the spherical SOFM on different data and the use of possible measures to create spherical SOFM graphical representations are discussed in

**3. Summary of previous work on core-data from the Neolithic Tell site at**

Coring, as a minimally destructive technique, facilitates the definition of subsurface units, provides a clear view of the buried surfaces on which occupations took place [23]. The mac‐ roscopic examination of all twelve cores drawn from the subsurface investigation conducted on the tell, revealed three basic stratigraphic units: bedrock, occupation deposits and a topsoil layer. Given, therefore, their relative macrostratigraphic similarity and the rather broad stratigraphic resolution/delineation required from the cores at Paliambela, three cores (out

marks, respectively.

**2. Spherical self-organizing feature map**

some physical aspect of the n-dimensional data.

Sangole [17] and Sangole and Knopf [18].

**Paliambela (Pieria region-Northern Greece-Fig.1)**

**Figure 1.** Map of Greece, showing the location of Paliambela (Source:'Paliambela excavation' archive).

The use of the spherical-SOFM non-linear method was also able to recognize and to provide a visual representation of micro-artefact patterns in archaeological contexts (i.e., a colluvial deposit from a Hellenistic Theatre in NW Greece) affected *only* by natural formation proc‐ esses [10, 11].

The implication of the applied non-linear method (i.e., spherical-SOFM) is that it has the 'ability' to demonstrate the dynamics of cultural or natural formation processes in leaving non-linear 'signals' in archaeological contexts being in a *'non equilibrium'* state until the time of recovery. Therefore, the rationality for developing such recognitions in archaeological contexts is to release the dynamics of formation processes since archaeological patterning is arguably (at least for the most part) the result of the interplay between many complex proc‐ esses, both cultural and non-cultural (natural) [2, 28, 1, 16]. Therefore, this type of recogni‐ tion is of critical importance also in core-data, since this type of data provide broader spatial information and are more sensitive in both cultural and natural formation processes.

Section 2 briefly describes how the spherical self-organizing map creates a 3D visual or graphical representation of the data. Section 3 presents a summary of previous geoarchaeo‐ logical work on core-data from the Neolithic Tell site at Paliambela (Pieria region-Northern Greece) while section 4 and section 5 offer the results of this study and some concluding re‐ marks, respectively.

## **2. Spherical self-organizing feature map**

characterized by different non-linear associations, further suggesting that these were possi‐

**Figure 1.** Map of Greece, showing the location of Paliambela (Source:'Paliambela excavation' archive).

esses [10, 11].

The use of the spherical-SOFM non-linear method was also able to recognize and to provide a visual representation of micro-artefact patterns in archaeological contexts (i.e., a colluvial deposit from a Hellenistic Theatre in NW Greece) affected *only* by natural formation proc‐

The implication of the applied non-linear method (i.e., spherical-SOFM) is that it has the 'ability' to demonstrate the dynamics of cultural or natural formation processes in leaving non-linear 'signals' in archaeological contexts being in a *'non equilibrium'* state until the time of recovery. Therefore, the rationality for developing such recognitions in archaeological contexts is to release the dynamics of formation processes since archaeological patterning is arguably (at least for the most part) the result of the interplay between many complex proc‐ esses, both cultural and non-cultural (natural) [2, 28, 1, 16]. Therefore, this type of recogni‐ tion is of critical importance also in core-data, since this type of data provide broader spatial

information and are more sensitive in both cultural and natural formation processes.

Section 2 briefly describes how the spherical self-organizing map creates a 3D visual or graphical representation of the data. Section 3 presents a summary of previous geoarchaeo‐ logical work on core-data from the Neolithic Tell site at Paliambela (Pieria region-Northern

bly formed through different cultural formation processes [8].

222 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

The Spherical Self-Organizing Feature Map (S-SOFM), introduced by Kohonen [7], maps ndimensional data into a low-dimensional space. The spherical SOFM [17] the low-dimen‐ sional space is a tessellated sphere that is formed by subdividing an icosahedron. Every vertex on this sphere is a strategic location of an n-dimensional vector that represents an en‐ semble of similar data vectors which are assigned to the vector during the mapping opera‐ tion. It is therefore necessary to visually enhance variations in the data using the physical attributes of the mapping lattice. The benefit of a spherical lattice in the implementation of the S-SOFM is that the enclosed space can be used to generate a 3D visual representation of some physical aspect of the n-dimensional data.

Conventional implementation of the S-SOFM method have used a 2D lattice as the low-dimen‐ sional space, and associations in the data are visualized by means of a terrain map, wherein el‐ evation represents some aspect of the vector(s) at that location [27, 24]. Relative similarity between data vectors mapped into the sphere can be visualized by introducing distortions in the sphere accompanied by changes in the colour. Informative characteristics of the data are re‐ flected as distortions and colour gradations on the surface of the sphere. The formulation of these measures is a non-trivial task and often application dependent. The measures reflect de‐ sired data correlations (either linear or non-linear) and must be defined by the researcher who is familiar with the underlying data set. It is this aspect of the S-SOFM that differs from existing literature about the self-organizing feature map. The S-SOFM utilizes the spherical lattice of the S-SOFM space to generate a visual form of the clustered data that is more intuitive and easy to perceive. A visual form of the data is created by scaling the radial distance of the vertices on the sphere in proportion to a measure characterizing some physical aspect of the data. Exam‐ ples illustrating the various implementations of the spherical SOFM on different data and the use of possible measures to create spherical SOFM graphical representations are discussed in Sangole [17] and Sangole and Knopf [18].

## **3. Summary of previous work on core-data from the Neolithic Tell site at Paliambela (Pieria region-Northern Greece-Fig.1)**

Coring, as a minimally destructive technique, facilitates the definition of subsurface units, provides a clear view of the buried surfaces on which occupations took place [23]. The mac‐ roscopic examination of all twelve cores drawn from the subsurface investigation conducted on the tell, revealed three basic stratigraphic units: bedrock, occupation deposits and a topsoil layer. Given, therefore, their relative macrostratigraphic similarity and the rather broad stratigraphic resolution/delineation required from the cores at Paliambela, three cores (out of 12) were selected for analysis (i.e., nos 1083-84-85, Fig.2). These were judged to provide good site coverage from east to west, and thus offer information regarding the depth and thickness of the cultural deposits on the tell, on a coarse temporal and spatial scale.

The analysis of three cores (out of twelve) defined the stratigraphy of the site on a coarse temporal and spatial scale. The culturally sterile bedrock was identified at ca 2m depth and the initial human occupation probably started on top of the bedrock, because remnants of any overlying palaeosol have not been recognised, suggesting that this might have been stripped or reworked by subsequent human activity.

The analysis of the occupation deposits revealed significant variation, both temporal and spatial, in micro-artefacts (burnt clay, unburnt clay, shell, bone and charcoal) and so, pre‐ sumably, in human activity on the site. This indicates that the surviving occupation deposits built up sufficiently rapidly and bury and preserve variable concentrations of micro-arte‐ facts. In short, the analysis of the cores revealed that the tell component of the site might have been the product of long term anthropogenic accretion of sediment and artefactual ma‐ terial that created a low mound, with 0.5-2m of *surviving* occupation debris. [8].

**Figure 2.** Contour map of the Neolithic tell showing the location of the Cores. Contours at 10m (Source: 'Paliambela

Non-Linear Spatial Patterning in Cultural Site Formation Processes

http://dx.doi.org/10.5772/51193

225

Micro-artefact density (*D*) was obtained by using the following equation: *D= m/v,* where *m* is the weight of each material class, and *v* is the volume of each sample. This method is rather simple but needs the total sub-sample to be sorted for micro-artefacts and the various mate‐

The construction of the S-SOFM graphical representation was based on a database of 120 five-dimensional records, each dimension representing a micro-artefact category. Every row represented the point-counting results. Figure 3 shows the formation of three distinct white regions that correspond to the micro-artefact core-data from the site. A non-linear structure lies within this statistical space which can be distinguished into three separate sub-struc‐ tures. The spherical-SOFM pattern recognition procedure provides a comprehensive pre‐ liminary visual representation of inherent non-linear characteristics in data, serving as the initial step in the analysis of the multidimensional micro-artefact data. In this study three meaningful components were revealed – which appeared to be the determinants for the con‐ stitution of the analysed data set. This further suggests that the three groups of contexts (i.e., micro-artefacts from the three cores) from the site were possibly formed through different

It is important to mention that the five classes of micro-artefacts set for analysis -i.e., microshell, micro-bone, microfragments of charcoal, microfragments of burnt clay, and microfrag‐ ments of unburnt clay- generate from an interpretatively complicated set of larger artefacts,

The preservation of such materials in an archaeological context indeed, is closely connected not only with the length of deposition but also with the rate and type of weathering [22].

rial types to be weighed in a high precision electronic balance.

those made of friable materials -the so called 'size unstable' [22].

excavation' archive).

formation processes.

## **4. Non-linear micro-artefact patterning as a general indicator of differences in cultural site formation processes**

### **4.1. Laboratory procedures**

The laboratory procedure followed in micro-artefact analysis used two divisions of the phi (ф) scale, that is -2.00ф and 0ф. Contents of the bulk samples were passed through a stack of 4mm (-2.00ф) and 1mm (0ф) sieves. The material retained in the 1mm sieve created the subsample that was processed for micro-artefacts and an optical microscope was used for iden‐ tifications. Five micro-artefact categories were identified in the deposits from the cores: unburnt clay (e.g., from mudbricks, wattle and daub constructions), burnt clay (i.e., burnt specimens of the previous category), shell (marine shells), bone (animal bone), and charcoal (charred organic particles).

Shell, bone and charcoal were easily distinguished, but the more problematic distinction be‐ tween unburnt clay and burnt clay was based on the following observations: unburnt clay grains were often very fragile even in this small fraction and were, in most cases, subdiscoi‐ dal, ranging in colour from light grey to dark grey; burnt clay fragments were, in most cases, spherical particles of brownish colour, and relatively more solid than the unburnt clay ones.

In the core samples (total number: 120 sediment samples, ca. 1000kg each), the total subsample was sorted for micro-artefacts and the total mass for each material class was weigh‐ ed on an electronic precision balance. Physically sorting the total sub-sample for microartefacts, followed by weighing of each category, provides a representative picture of the micro-artefacts present in a sample and is feasible when small sample sizes are involved in analysis, as in the case of the cores.

of 12) were selected for analysis (i.e., nos 1083-84-85, Fig.2). These were judged to provide good site coverage from east to west, and thus offer information regarding the depth and

The analysis of three cores (out of twelve) defined the stratigraphy of the site on a coarse temporal and spatial scale. The culturally sterile bedrock was identified at ca 2m depth and the initial human occupation probably started on top of the bedrock, because remnants of any overlying palaeosol have not been recognised, suggesting that this might have been

The analysis of the occupation deposits revealed significant variation, both temporal and spatial, in micro-artefacts (burnt clay, unburnt clay, shell, bone and charcoal) and so, pre‐ sumably, in human activity on the site. This indicates that the surviving occupation deposits built up sufficiently rapidly and bury and preserve variable concentrations of micro-arte‐ facts. In short, the analysis of the cores revealed that the tell component of the site might have been the product of long term anthropogenic accretion of sediment and artefactual ma‐

The laboratory procedure followed in micro-artefact analysis used two divisions of the phi (ф) scale, that is -2.00ф and 0ф. Contents of the bulk samples were passed through a stack of 4mm (-2.00ф) and 1mm (0ф) sieves. The material retained in the 1mm sieve created the subsample that was processed for micro-artefacts and an optical microscope was used for iden‐ tifications. Five micro-artefact categories were identified in the deposits from the cores: unburnt clay (e.g., from mudbricks, wattle and daub constructions), burnt clay (i.e., burnt specimens of the previous category), shell (marine shells), bone (animal bone), and charcoal

Shell, bone and charcoal were easily distinguished, but the more problematic distinction be‐ tween unburnt clay and burnt clay was based on the following observations: unburnt clay grains were often very fragile even in this small fraction and were, in most cases, subdiscoi‐ dal, ranging in colour from light grey to dark grey; burnt clay fragments were, in most cases, spherical particles of brownish colour, and relatively more solid than the unburnt clay ones.

In the core samples (total number: 120 sediment samples, ca. 1000kg each), the total subsample was sorted for micro-artefacts and the total mass for each material class was weigh‐ ed on an electronic precision balance. Physically sorting the total sub-sample for microartefacts, followed by weighing of each category, provides a representative picture of the micro-artefacts present in a sample and is feasible when small sample sizes are involved in

thickness of the cultural deposits on the tell, on a coarse temporal and spatial scale.

terial that created a low mound, with 0.5-2m of *surviving* occupation debris. [8].

**4. Non-linear micro-artefact patterning as a general indicator of**

stripped or reworked by subsequent human activity.

224 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

**differences in cultural site formation processes**

**4.1. Laboratory procedures**

(charred organic particles).

analysis, as in the case of the cores.

**Figure 2.** Contour map of the Neolithic tell showing the location of the Cores. Contours at 10m (Source: 'Paliambela excavation' archive).

Micro-artefact density (*D*) was obtained by using the following equation: *D= m/v,* where *m* is the weight of each material class, and *v* is the volume of each sample. This method is rather simple but needs the total sub-sample to be sorted for micro-artefacts and the various mate‐ rial types to be weighed in a high precision electronic balance.

The construction of the S-SOFM graphical representation was based on a database of 120 five-dimensional records, each dimension representing a micro-artefact category. Every row represented the point-counting results. Figure 3 shows the formation of three distinct white regions that correspond to the micro-artefact core-data from the site. A non-linear structure lies within this statistical space which can be distinguished into three separate sub-struc‐ tures. The spherical-SOFM pattern recognition procedure provides a comprehensive pre‐ liminary visual representation of inherent non-linear characteristics in data, serving as the initial step in the analysis of the multidimensional micro-artefact data. In this study three meaningful components were revealed – which appeared to be the determinants for the con‐ stitution of the analysed data set. This further suggests that the three groups of contexts (i.e., micro-artefacts from the three cores) from the site were possibly formed through different formation processes.

It is important to mention that the five classes of micro-artefacts set for analysis -i.e., microshell, micro-bone, microfragments of charcoal, microfragments of burnt clay, and microfrag‐ ments of unburnt clay- generate from an interpretatively complicated set of larger artefacts, those made of friable materials -the so called 'size unstable' [22].

The preservation of such materials in an archaeological context indeed, is closely connected not only with the length of deposition but also with the rate and type of weathering [22]. Moreover, micro-artefacts provide different information than do larger artefacts and defi‐ nitely should not be used simply to reflect 'noise' in larger artefacts [6].

arguably reflect long-term continuity of distinct patterns of spatial organisation of behav‐ iour. The term 'continuity' is conceptualized here as the cultural product of different social systems (Neolithic or later) that inhabited Paliambela. Their cultural outcomes, embedded in and decoded from the archaeological sediments, contributed significantly to the site's for‐ mation, transforming it into a cultural product. Therefore, the archaeological sediments of Paliambela, enclose significant cultural information, and this study has demonstrated the

Non-Linear Spatial Patterning in Cultural Site Formation Processes

http://dx.doi.org/10.5772/51193

227

**5. Conclusions: The importance of understanding formation processes in**

Since Schiffer's [19] original recognition of the importance of studying and understanding the formation processes of the archaeological record, many authors have pointed out their critical importance [5, 12, 13, 25]. Moreover, it is now widely accepted that variability is in‐ troduced into the archaeological record through cultural and non-cultural formation proc‐

The unit of analysis appropriate for identifying formation processes is, according to Schiffer [20] the deposit, but "viewing the deposit as a single discrete depositional event or process has its problems, as a single depositional process can give rise to materials in different de‐ posits, and conversely, a single deposit can contain the products of many different deposi‐

However, despite the recognised importance of cultural and natural processes in the forma‐ tion of the archaeological record, studies addressing the interpretative potential of micro-ar‐ tefacts remain relatively limited, although micro-artefacts, due to their abundance and incorporation in an archaeological deposit constitute a significant part of the cultural parti‐ cles present and may provide information on the cultural and natural formation processes

Dunnel and Stein [6] outline some of the important characteristics of micro-artefacts that compel their consideration as archaeological data of the first order. They note, that informa‐ tion content may be different for micro-artefacts than for larger artefacts and they may be most informative about different things (e.g., particle transport and site formation process‐ es). Equally important, processes that generate microscopic artefacts vary depending on ma‐ terial and context [6]. These last two issues, differing information content and differing formation processes within the micro-scale are important reasons for undertaking micro-ar‐

Then again, attempting to define cultural and natural formation processes in a site focusing, for example, either in their variability or in the proportional correlation among micro-arte‐ fact classes may be misleading because their archaeological significance rests upon under‐ standing the interaction among, the almost, numerous variables within a sequence which

esses which distort systemic patterns as well as creating their own patterns [20].

potential of the non-linear method to help identify this information.

**a non-linear world**

tional processes" [20].

tefact analysis [6].

occurring in a deposit [4, 3, 26, 6, 21].

would determine their transport potential.

Therefore, the researcher cannot assume that, for example, chronologically distant archaeo‐ logical contexts will provide similar or different micro-artefact patterning due to the many factors that may account for the observed pattern. The implication is that it enhances at‐ tempts for developing interpretations on micro-artefact patterning by providing strong pat‐ tern recognition.

The observation of this pattern in cultural indicators such as micro-artefacts should be related at least in part (and arguably for the most part) with differences in the spatial organization of activities carried out in the site and ending up in the deposits. In other words, it should be relat‐ ed with spatial differences in cultural formation processes. That these differences in cultural processes had become so embedded in the sedimentary traces of the deposits arguably reflects *long-term continuity* of distinct patterns of spatial organisation of behaviour.

In this study, the critical prerequisite was rather the depiction of the existence of differences in formation processes (and arguably of cultural formation processes) and of spatial content between different contexts (i.e., cores) from the site, on a broad spatial and temporal scale, than a detailed presentation of the spatial and temporal use of space on the tell settlement or of differences in formation processes between different activity areas across the site.

Despite the natural and cultural agents/processes that have disorganized the site's behav‐ ioural contexts, the archaeological sediments from Paliambela still preserve significant nonlinear behavioural information. The spatial differences in cultural formation processes arguably reflect long-term continuity of distinct patterns of spatial organisation of behav‐ iour. The term 'continuity' is conceptualized here as the cultural product of different social systems (Neolithic or later) that inhabited Paliambela. Their cultural outcomes, embedded in and decoded from the archaeological sediments, contributed significantly to the site's for‐ mation, transforming it into a cultural product. Therefore, the archaeological sediments of Paliambela, enclose significant cultural information, and this study has demonstrated the potential of the non-linear method to help identify this information.

Moreover, micro-artefacts provide different information than do larger artefacts and defi‐

Therefore, the researcher cannot assume that, for example, chronologically distant archaeo‐ logical contexts will provide similar or different micro-artefact patterning due to the many factors that may account for the observed pattern. The implication is that it enhances at‐ tempts for developing interpretations on micro-artefact patterning by providing strong pat‐

The observation of this pattern in cultural indicators such as micro-artefacts should be related at least in part (and arguably for the most part) with differences in the spatial organization of activities carried out in the site and ending up in the deposits. In other words, it should be relat‐ ed with spatial differences in cultural formation processes. That these differences in cultural processes had become so embedded in the sedimentary traces of the deposits arguably reflects

**Figure 3.** View of the S-SOFM graphical representation showing the formation of three distinct white regions – corre‐

In this study, the critical prerequisite was rather the depiction of the existence of differences in formation processes (and arguably of cultural formation processes) and of spatial content between different contexts (i.e., cores) from the site, on a broad spatial and temporal scale, than a detailed presentation of the spatial and temporal use of space on the tell settlement or

Despite the natural and cultural agents/processes that have disorganized the site's behav‐ ioural contexts, the archaeological sediments from Paliambela still preserve significant nonlinear behavioural information. The spatial differences in cultural formation processes

of differences in formation processes between different activity areas across the site.

sponding to micro-artefact core-data (each for a core).

nitely should not be used simply to reflect 'noise' in larger artefacts [6].

*long-term continuity* of distinct patterns of spatial organisation of behaviour.

tern recognition.

226 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

## **5. Conclusions: The importance of understanding formation processes in a non-linear world**

Since Schiffer's [19] original recognition of the importance of studying and understanding the formation processes of the archaeological record, many authors have pointed out their critical importance [5, 12, 13, 25]. Moreover, it is now widely accepted that variability is in‐ troduced into the archaeological record through cultural and non-cultural formation proc‐ esses which distort systemic patterns as well as creating their own patterns [20].

The unit of analysis appropriate for identifying formation processes is, according to Schiffer [20] the deposit, but "viewing the deposit as a single discrete depositional event or process has its problems, as a single depositional process can give rise to materials in different de‐ posits, and conversely, a single deposit can contain the products of many different deposi‐ tional processes" [20].

However, despite the recognised importance of cultural and natural processes in the forma‐ tion of the archaeological record, studies addressing the interpretative potential of micro-ar‐ tefacts remain relatively limited, although micro-artefacts, due to their abundance and incorporation in an archaeological deposit constitute a significant part of the cultural parti‐ cles present and may provide information on the cultural and natural formation processes occurring in a deposit [4, 3, 26, 6, 21].

Dunnel and Stein [6] outline some of the important characteristics of micro-artefacts that compel their consideration as archaeological data of the first order. They note, that informa‐ tion content may be different for micro-artefacts than for larger artefacts and they may be most informative about different things (e.g., particle transport and site formation process‐ es). Equally important, processes that generate microscopic artefacts vary depending on ma‐ terial and context [6]. These last two issues, differing information content and differing formation processes within the micro-scale are important reasons for undertaking micro-ar‐ tefact analysis [6].

Then again, attempting to define cultural and natural formation processes in a site focusing, for example, either in their variability or in the proportional correlation among micro-arte‐ fact classes may be misleading because their archaeological significance rests upon under‐ standing the interaction among, the almost, numerous variables within a sequence which would determine their transport potential.

The example offered in this study indicates that similar types of micro-artefacts within dif‐ ferent archaeological contexts across the site exhibit significant non-linear information plau‐ sibly as a result of different types of formation processes that were assumed to imply, for the most part, differences in cultural formation processes. In any case, stronger interpretation can only be achieved by strong micro-artefact pattern recognition [9, 10] especially in cases of archaeological deposits sensitive to cultural formation processes.

[5] Goldberg, P., Nash, D. T., & Petraglia, M. D. (1993). Press Formation Processes. *Ar‐*

Non-Linear Spatial Patterning in Cultural Site Formation Processes

http://dx.doi.org/10.5772/51193

229

[6] Dunnell, R. C., & Stein, J. K. (1989). Theoretical issues in the interpretation of micro‐

[7] Kohonen, T. Self-organized formation of topologically correct feature maps. *Biological*

[8] Kontogiorgos, D. (2008). Geoarchaeological and microartifacts analysis of archaeo‐ logical sediments. *A case study from a Neolithic Tell site in Greece. Nova Science Publish‐*

[9] Kontogiorgos, D., Leontitsis, A., & Sangole, A. (2007). Telling a non linear story: The investigation of microartefacts non linear structure. *Journal of Archaeological Science*,

[10] Kontogiorgos, D., & Preka, K. (2009). From Neolithic to Hellenistic. A Geoarchaeo‐ logical Approach to the Burial of the a Hellenistic Theatre: The Evidence from Parti‐ cle Size Analysis and Microartifacts. *On Site Geoarchaeology on a Neolithic Tell Site in Greece: Archaeological Sediments,Microartifacts and Softwear Development, Kontogiorgos,*

[11] Kontogiorgos, D., & Leontitsis, A. (2011). Is it Visible? *Micro-artefacts Non-Linear Stuc‐ ture and Natural Formation Processes In: Self-Organizing Maps and Novel Algorithm De‐*

[12] Mc Guire, R. H. (1995). Behavioural Archaeology: Reflections of a Prodigal Son. *J. M. Skibo, W. H.Walker and A. E. Nielsen (eds.) Expanding Archaeology*, 162-177, Salt Lake

[13] Reid, J. J. (1995). Four Strategies after Twenty Years: A Return to Basics. *J. M. Skibo, W. H. Walker and A. E. Nielsen (eds.) Expanding Archaeology*, 15-21, Salt Lake City, Uni‐

[14] Rosen, A.M. (1986). *Cities of Clay: The Geoarchaeology of Tells.University of Chicago*

[15] Rosen, A. M. (1989). Ancient Town and City Sites: A View from the Microscope.

[16] Ormerod, P. (2005). *Why most things fail: Evolution, extinction, and economics*, London,

[17] Sangole, A. (2003). Data-driven modeling using spherical self-organizing feature maps. *PhD thesis, University of Western Ontario*, Canada, Universal Publishers,

[18] Sangole, A., & Knopf, G. K. (2003). Visualization of random ordered numeric data

sets using self-organized feature maps. *Computers and Graphics*, 963-976.

*chaeological Context Madison, Wisconsin: Prehistory*.

*D. Ed., Nova Science Publishers, Inc., New York*, 71-80.

*sign. Intech open access Publishers. Vienna, Austria.*, 643-648.

artifacts. *Geoarchaeology*, 31-42.

City, University of Utah Press.

versity of Utah Press.

*American Antiquity*, 564-578.

*Press, Chicago*.

Faber & Faber.

1-58112-319-1.

*Cybernetics*, 198(43), 59-69.

*ers Inc., New York*.

1532-1536.

Without underestimating the effects of natural processes or rather 'naively' expecting cul‐ tural factors to account for all the extant variability in an archaeological site, it seems that drawing logical connections between geoarchaeological data and past human activities up‐ grades and enhances cultural interference upon natural factors in a site's formation. The study of micro-artefacts, those cultural particles included into archaeological sediments, al‐ though by no means conclusive, can be utilised to identify forms of behaviour enacted with‐ in a site, when strong pattern recognition has been achieved [8].

New ways of describing differences in archaeological assemblages could only be effective if we could connect them with past human behaviour in a non static physical environment. The identification of cultural formation processes, spatially and temporally, on the micro-level in a complex site, as Paliambela, indicates that such discrimination is possible through the applica‐ tion of a certain methodology. More importantly, it calls for awareness of the multiplicity of scales at which these cultural processes can be traced. Although not exhaustive, the non linear spherical self organizing feature map method has provided a higher resolution with which to view the archaeological information encoded within archaeological sediments.

## **Author details**

Dimitris Kontogiorgos

Wiener Laboratory, American School of Classical Studies at Athens, Athens, Greece

## **References**


[5] Goldberg, P., Nash, D. T., & Petraglia, M. D. (1993). Press Formation Processes. *Ar‐ chaeological Context Madison, Wisconsin: Prehistory*.

The example offered in this study indicates that similar types of micro-artefacts within dif‐ ferent archaeological contexts across the site exhibit significant non-linear information plau‐ sibly as a result of different types of formation processes that were assumed to imply, for the most part, differences in cultural formation processes. In any case, stronger interpretation can only be achieved by strong micro-artefact pattern recognition [9, 10] especially in cases

Without underestimating the effects of natural processes or rather 'naively' expecting cul‐ tural factors to account for all the extant variability in an archaeological site, it seems that drawing logical connections between geoarchaeological data and past human activities up‐ grades and enhances cultural interference upon natural factors in a site's formation. The study of micro-artefacts, those cultural particles included into archaeological sediments, al‐ though by no means conclusive, can be utilised to identify forms of behaviour enacted with‐

New ways of describing differences in archaeological assemblages could only be effective if we could connect them with past human behaviour in a non static physical environment. The identification of cultural formation processes, spatially and temporally, on the micro-level in a complex site, as Paliambela, indicates that such discrimination is possible through the applica‐ tion of a certain methodology. More importantly, it calls for awareness of the multiplicity of scales at which these cultural processes can be traced. Although not exhaustive, the non linear spherical self organizing feature map method has provided a higher resolution with which to

view the archaeological information encoded within archaeological sediments.

Wiener Laboratory, American School of Classical Studies at Athens, Athens, Greece

[1] Ball, P. (2004). Critical mass: How one thing leads to another. Portsmouth, NH: Hei‐

[2] Bentley, R. A., & Maschner, H. D. G. (2003). Complex systems and archaeology. *Salt*

[3] Fladmark, K. R. (1982). Microdebitage analysis: initial considerations. *Journal of Ar‐*

[4] Hassan, F.A. (1978). Sediments in archaeology: Methods and implications for palae‐

oenvironmental and cultural analysis. *Journal of Field Archaeology*, 197-213.

of archaeological deposits sensitive to cultural formation processes.

228 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

in a site, when strong pattern recognition has been achieved [8].

**Author details**

**References**

Dimitris Kontogiorgos

nemann.

*Lake City: University of Utah Press*.

*chaeological Science*, 9, 205-220.


[19] Schiffer, M. B. (1972). Archaeological Context and Systemic Context. *American Antiq‐ uity*, 156-165.

**Chapter 12**

**Spatial Clustering Using Hierarchical SOM**

The amount of available geospatial data increases every day, placing additional pressure on existing analysis tools. Most of these tools were developed for a data poor environment and thus rarely address concerns of efficiency, high-dimensionality and automatic exploration [1]. Recent technological innovations have dramatically increased the availability of data on location and spatial characterization, fostering the proliferation of huge geospatial databas‐ es. To make the most of this wealth of data we need powerful knowledge discovery tools, but we also need to consider the particular nature of geospatial data. This context has raised new research challenges and difficulties on the analysis of multidimensional geo-referenced data. The availability of methods able to perform "intelligent" data reduction on vast amounts of high dimensional data is a central issue in Geographic Information Science

The field of knowledge discovery constitutes one of the most relevant stakes in GISc re‐ search to develop tools able to deal with "intelligent" data reduction [2, 3] and tame com‐ plexity. More than prediction tools, we need to develop exploratory tools which enable an

The term cluster analysis encompasses a wide group of algorithms (for a comprehensive re‐ view see [5]). The main goal of such algorithms is to organize data into meaningful struc‐ tures. This is achieved through the arrangement of data observations into groups based on similarity. These methods have been extensively applied in different research areas includ‐ ing data mining [6, 7], pattern recognition [8, 9], and statistical data analysis [10]. GISc has also relied heavily on clustering algorithms [11, 12]. Research on geodemographics [13-16], identification of deprived areas [17], and social services provision [18] are examples of the

> © 2012 Henriques et al.; licensee InTech. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

© 2012 Henriques et al.; licensee InTech. This is a paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

relevance that clustering algorithms have within today's GISc research.

Roberto Henriques, Victor Lobo and

Additional information is available at the end of the chapter

Fernando Bação

**1. Introduction**

http://dx.doi.org/10.5772/51159

(GISc) current research agenda.

improved understanding of the available data [4].


**Chapter 12**
