The variations in the technological and typological composition of lithic assemblages from the Middle Paleolithic remains a subject of debate. The knapping methods, site occupation patterns or the organization of lithic technology are some of the underlying factors commonly cited as causes of this variability. Adding up to this is the need to increase the possible methodological tools for the analysis and comparison of lithic
assemblages.
The following PhD addresses this problematic through the development of controlled
experimentations that allow to test and propose new analytical methods, and through the
analysis of four lithic assemblages belonging to two different Middle Paleolithic site of
the Iberian Peninsula (the levels VII and XIf of Esquilleu cave and the levels IIIa and IIIb
of Teixoneres Cave).
The work has been divided in four blocks. The first block presents the introduction and
the conceptual frameworks under which the approach has been undertaken. The second
block is subdivided in the methodology employed and the experimentations linked to de
development of the work. The third block presents the archaeological sites and levels and
the results from the analysis of the lithic assemblages. The fourth and last block presents
the discussion of the results (focusing on the importance of thee results, linking them to
other published results, pointing limitations and making suggestions for further research)
and the conclusions.
The first experimentation employs a set of experimental flaks with different proportions
of retouched along with a protocol for extracting measures from digital photographs. Six
different observers are asked to take digital photographs and use the protocol to obtain
measures of length of the edge and length of the retouched edge, allowing to compare the
deviation in the obtained values. The second experimentation employs a sample of 24
flakes that are successively retouched and following three different patterns of retouch.
For each episode the height of retouch, the length of retouched edge and lost weight are
recorded, allowing to determine the degree of correlation between the combination of two
measures (height of retouch and length of retouched edge) with the weight lost by retouch.
The third experimentation uses 451 flaking products of 33 different cores from five
different knapping methods (one of them is subdivided in two phases making a total of six categories). A series of measures and attributes common to lithic analysis are taken
and employed to build a decision tree (type of machine learning process).
Results from the first experimentation show that the use of digital photographs to obtain
measures from the edge (length of the edge, length of the retouched edge and percentage
of retouched edge) have a high degree of reproducibility. The second experimentation
allows to propose a new index (the AvtL) that allows to estimate the weight/volume lost
by retouch. Correlation is high and it can be applied to different patterns of retouch. The
third experimentation allows to generate a Machine Learning decision tree model to
identify knapping methods in flaking products. This last experimentation also yields
results about the reproducibility in the identification of knapping methods, the
directionality of confusions, and which variables characterize the products from different
knapping methods.
Results from the archaeological assemblages are structured in five sections for each
assemblage. The first section analyzes flakes, presenting the results from the raw material
analysis, raw material quality, attributed knapping methods, identified technological
products, dimensional analysis, percussion platform analysis, scar directions, transversal
sections and edge angles. The second section repeats this analysis on retouched artifacts
adding the typological composition. The third section compares the results from the
analysis of flakes and tools (comparing raw material composition, dimensions, etc.).
Fourth section analyses the intensity and extension of retouch according to raw materials,
attributed knapping methods and technological products. The fifth and last sections use a
K-means clustering method and values from intensity and extension of retouch along with
raw material quality values to stablish groups of tools. This allows to identify groups
according to the retouch intensity, and identify patterns of selection and possible curated
toolkits among the retouched artifacts.
The comparison between levels is made following two approaches. First, a factorial
analysis using the frequency of technological products and attribution of knapping
methods is employed to determine what is characterizing each complete assemblage and
their subsets of tools and flakes. The values of the factorial analysis are employed to
reduce the dimensionality of the data and as values of a dendogram which allows to
stablish relations of similarity between assemblages. The second approach uses average
values of intensity an extension of retouch, quality of raw materials, their modal distribution and size differences between retouched and unretouched products to stablish
differences between assemblages.
Results from level VII of Esquilleu Cave show an increase in the use of silified lutites
although quartzite’s are still the most employed raw material. Levallois, hierarchical
Discoid and Discoid are the most important knapping methods. retuch is more intense
and extended in quartzite’s, and cluster analysis show a preference for backed flakes,
cortical products and some Levallois flakes.
Results from level XIf show a dominance of the Quina knapping method and the
quartzite’s as raw material (however there is a higher variability of quartzite’s). Intensity
and extension of retouch is much higher than previous levels, but not all materials are
intensively retouched. There is an important relation between tool type and retouch
intensity.
Results from level IIIa are conditioned by the low number of materials which show a
marked duality between the catchment of local low quality raw materials (quartz and
limestone) and on local high/medium quality raw materials (mainly flint). Levallois and
Tranche de Saucisson are the main methods. Only one element can be considered as a
curated toolkit while the rest of retouched products can be considered as complementary
provisioning strategies.
Results from level IIIb show a decrease of the duality between local/low quality and nonlocal/
high quality materials in favor of the last ones. This is a result of the catchment of a
higher diversity of raw materials (sandstones, quartzites, hornfels, etc.) which results in a
higher average raw material quality. Maine knapping methods continue to be Levallois
and Tranche de Saucisson. Analysis of intensity of retouch show important differences
respect to the previous levels. Distribution of retouch intensity is more gradual and several
products can be considered as curated toolkits.
The comparison between levels shows that XIf is characterized by the exclusivity of the
attribution of the Quina knapping method along with its own technological products (such
as the platform wrapping flakes or the Quina resharpening flakes). This makes the Quina
assemblages the most differentiated of all assemblages. Levels IIIa and IIIb are
differentiated from level VII because of the important presence of cortical core edge
flakes, the attribution of the Tranche de Saucisson method, and the low attribution of the
Discoide method and low presence of products such as pseudo-Levallois points. The
comparison following retouch intensity and extension, raw material quality and size differences between tools and flakes show important differences between levels. Despite
this levels XIf and IIIb are more similar than levels VII and IIIa which can be linked to
the intensity of occupations.