Ethno-geochemical and phytolith studies of activity related patterns: A case study from Al Ma'tan, Jordan

Understanding Neolithic sites in southwest Asia is often difficult because of the lack of preservation of organic remains and the effects of various taphonomic processes that alter the original record. Here, we use an ethnographic approach to test the potential of using plant phytoliths and geochemistry to aid our interpretation of southwest Asian Neolithic sites. Our study of a recently abandoned stone and mud constructed village in Jordan, shows that for certain activity types, phytoliths and geochemistry can help distinguish different construction methods and functions, particularly for burnt areas, animal use areas and where there has been the addition of a specific construction material. For features constructed from the same source materials distinctions are more problematic. Geochemical and phytolith proxies were individually effective in distinguishing activity areas and construction materials, but signals were diminished when the statistical analysis was run on both forms of evidence combined. It is therefore recommended that the data from plant phytolith and geochemical analyses are subject to separate statistical tests and that the two sets of results are used in combination to interpret archaeological sites and their uses. the diverse nature of potential activities; the overlapping of elemental signatures for numerous activities; post-depositional mixing of sediments; post-depositional leaching; differential leaching of elements; and post-depositional alterations.


Introduction
The Neolithic in Southwest Asia The Neolithic in southwest Asia (c 11,700-7800 cal BP) is an important period in human history that saw the advent of sedentism, agriculture and ultimately paved the way for increased social complexity and urbanism. It is also, however, one of the most poorly understood. There are many factors that limit our understanding of this important period. One is the paucity of Neolithic sites and, often when sites are found, preservation is poor, particularly for organic remains. A further consideration for all archaeological investigations is that even when preservation is good, interpreting the evidence can be problematic because many variables influence the archaeological record and impede understanding. For example, the diverse nature of potential activities; specific site formation processes (e.g. cleaning prior to abandonment or periods of disuse); overlapping signatures for numerous activities; post-depositional mixing of sediments; post-depositional (and differential) leaching; and post-depositional alterations are just some of the processes that affect and alter the archaeological record. Furthermore, archaeologists frequently do not understand what types of evidence and signals can be produced from different activities (see Shillito 2017 for a full discussion).
In the past artefact patterning has been used to interpret past activities, however artefacts rarely represent specific in situ actions because they may have been removed, purposefully placed or lost within contexts unrelated to their original use. Only a fraction, if any, of the material record of activities may remain in the archaeological record (Shillito 2017). An example of this is from the analysis of site activity areas in a Viking Age house in Iceland, where a spindle whorl was recovered from hearth deposits, clearly not representing in situ activity (Milek and Roberts 2013). Artefact distributions are also often the result of cleaning, abandonment or trampling (e.g. Çatalhöyük, Shillito 2017, 9). There has also been significant debate as to whether 'activities' can be detected or whether results are more representative of construction materials. Artefact and microremains can reflect other processes such as materials embedded within the floors resulting from construction rather than representing a specific activity on the floor itself (Tsartsidou et al. 2009;Shillito 2017). As a result, many archaeological sites, particularly form earlier time periods, are comprised of a series of structures, the form and function of which is difficult, if not impossible, to interpret.
In order to help address this problem, we took an ethnographic approach to further our understanding of how everyday activities leave microscopic or invisible traces. Human behaviour rarely produces a discreet signal and many different lines of evidence have been used to investigate activities and use of space (Shillito 2017). We selected a dual proxy approach, namely phytoliths and geochemical residues which, because they are inorganic, are more likely to survive in and be recovered from, the archaeological record compared with other evidence such as macrobotanical and faunal remains, making them a valuable source of information. Furthermore, as these remains are microscopic or invisible it is more probable that they will remain in the sediments in comparison to macroscopic artefacts that could have been removed or cleaned away. This analysis will be supported by targeted micromrophological sampling which will be the subject of a forthcoming paper.

Past Use of Phytolith and Geochemical Data in Ethnoarchaeology
Phytoliths Phytolith analysis has frequently been employed in ethnoarchaeological research to address a wide range of research questions. Phytoliths have been investigated in modern contexts to: locate animal pens (Shahack-Gross, Marshall, and Weiner 2003;Shahack-Gross et al. 2004); examine domestic activities (Portillo et al. 2014); identify irrigation (Rosen and Weiner 1994;Madella et al. 2009;Jenkins, Jamjoum, and Al Nuimat 2011;Weisskopf et al. 2014;Jenkins et al. 2016); examine Bedouin tents (Jenkins, Baker, and Elliott 2011); further our understanding of cooking installations (Gur-Arieh et al. 2013); and to provide crop processing models (Harvey and Fuller 2005).
More pertinent for this study, Tsartsidou et al. (2008Tsartsidou et al. ( , 2009 conducted an ethnographic phytolith analysis of the use of space in a village in northern Greece with good effect. The results from these studies were then used to interpret phytolith assemblages from the Neolithic site of Makri, Greece and it was concluded that Markri was a permanently occupied site with a mixed agricultural and pastoral economy (Tsartsidou et al. 2009).

Geochemistry
Human habitation can significantly affect the chemical soil composition leading to enrichments and depletions of specific chemical elements and the formation of archaeological soils (Oonk, Slomp, and Huisman 2009). As such, elemental analysis has frequently been used to examine the use of space and activity areas in ethnoarchaeological and archaeological contexts (e.g. Middleton and Price 1996;Hutson and Terry 2006;Holliday and Gartner 2007;Wilson, Davidson, and Cresser 2009).
Considerable ethnographic research on how geochemistry provides information about the use of space has been undertaken by researchers working in Mexico. Extensive studies by Barba, Manzanilla and colleagues demonstrated that it was possible to identify different activity areas through changes in chemical concentrations with: rest/sleeping areas and thoroughfares being depleted in chemical compounds; food preparation areas having low phosphate values; and food consumption and animal penning areas having high phosphate values. They then used these results to interpret a number of archaeological sites in Mexico (Barba and Bello 1978;Barba and Denis 1981;Mejia and Barba 1988;Manzanilla and Barba 1990). Similarly, Middleton and Price (1996) analysed a range of floors from ethnographic and archaeological sites in Mexico. They demonstrated that floors could be identified on the basis of their geochemical signature and applied this method successfully to samples from two different archaeological sites ranging in age from 4000 BP to 800 AD (Middleton and Price 1996), while Smyth (1990) found increased concentrations of phosphorous (P) and calcium (Ca) in areas of maize treatment and preparation in Maya settlements in the Puuc region of Mexico.
However, there has been considerable debate and criticism of geochemical patterning studies to successfully distinguish the use of space and specific activities (e.g. Oonk, Slomp, and Huisman 2009;Canti and Huisman 2015, 100-101). Criticisms of this approach are based upon the numerous possibilities for each elemental signature e.g.: the diverse nature of potential activities; the overlapping of elemental signatures for numerous activities; post-depositional mixing of sediments; post-depositional leaching; differential leaching of elements; and post-depositional alterations.

Combined Approaches
Increasingly archaeologists are turning to combined approaches, or a multi-proxy methodology, to help understand the use of space. Shahack-Gross, Marshall, and Weiner (2003) used a combined phytolith and geochemical approach to study penning deposits from one occupied and four abandoned Masai settlements in southern Kenya. They found that the density of phytoliths was higher in penning deposits, while the geochemical analysis demonstrated that penning areas were distinct from regional sediments and from features inside the settlements such as hearths (Shahack-Gross et al. 2004). Rondelli et al. (2014; see also Madella et al. 2014) used geoarchaeology to look at different activity areas within an ethnographic site in North Gujarat, India, to determine if certain activities left specific signatures. They found that it was possible to distinguish food storage areas through high protein levels, food preparation areas through the presence of fatty acids and areas of burning and fuel use through geochemistry. Notably, their results found that fire installations where dung had been used as the fuel source, had higher values of aluminium (Al), barium (Ba), calcium (Ca), cobalt (Co), chromium (Cr), iron (Fe), manganese (Mn), molybdenum (Mo), nickel (Ni) and lead (Pb), while fire installations where wood had been used as the fuel source had higher values of Ca, potassium (K), magnesium (Mg), Al and P Rondelli et al. 2014).
Ethnoarchaeological research involving phytoliths and/or geochemistry is becoming increasingly prevalent and a full review of geo-ethnoarchaeological methods, can be found in Friesem (2016). The body of research focused on this area demonstrates the value of ethnographic analysis as a means to further our understanding of how certain activities can leave specific anthropogenic signatures that can help us interpret these activities archaeologically. However, while these studies show great promise, the types of structures and settlements which were the focus of these earlier studies are not directly comparable to southwest Asian Neolithic sites, a time period and region which is critical for our understanding of the development from mobile hunter-gatherers to sedentary farmers.

The Ethnographic Village of Al Ma'tan
To address this problem, we conducted an ethnoarchaeological study in Jordan which could be used as a direct comparison for southwest Asian Neolithic sites. The abandoned village of Al Ma'tan was chosen because it was constructed using similar building materials and methods to southwest Asian Neolithic sites. Al Ma'tan is located in the At Tafila governorate of Jordan ( Figure 1) and was founded in the late nineteenth or early twentieth centuries, with some reported variation as to the exact date of establishment. The village was slowly abandoned from the 1960s onwards, with the final inhabitants leaving in the 1980s, though some houses continue to be used for animal housing to the present day. The community resettled close-by in a modern village, now town, next to a major highway which provides better infrastructure and links to the capital city of Amman.
When Al Ma'tan was occupied, the inhabitants were semi-nomadic, practising rainfed arable farming, sheep/goat pastoralism and tree cultivation in adjacent orchards irrigated by local springs. The houses were most intensively lived-in during the winter, with goat hair tents used in the warmer months while out roaming in the landscape. Houses were important for yearround storage of food, fodder, agricultural equipment and household goods. In winter, the goat hair tents were also stored inside.
The architecture of the houses at Al Ma'tan is typical of southern Jordan and has been described by, for example, Khammash (1986), Biewers (1997) and, most recently, Twaissi, Abuhalaleh, and Abudanah (2016). The houses at Al Ma'tan are generally singleroomed, all one-storied and have flat roofs. Each one has a single entry door, usually with small openings above for ventilation. The most distinctive internal architectural feature is the use of arches which support the roof beams. Bays between the walls from which these arches spring are commonly used as 'rooms' for sleeping and for storage. The houses are fondly remembered by their former residents as very practical and warm in winter as well as cool in summer, but the use of natural materials meant that they required constant maintenance. Later in the village's history the residents started to use modern materials such as concrete, but most of the village is built from traditional materials.
From ethnographic discussions with the former Al Ma'tan residents, we know the village was built using local stone, mud and plant materials. One particularly important clay used in construction at Al Ma'tan is regionally referred to as samaga-an unaltered natural 'clay' collected from known sources around the village. Technically samaga is comprised of clay sized particles and not clay minerals (smectites) and the samaga collected and used at Al Ma'tan has a high carbonate content (pers. comm. Dr Fathi Shaqour, 10/5/2016). Samaga can be found in the At Tafila area in three colour variations: red, yellow and blue/green. The colour of the clay is dependent on the amount of iron (Fe) oxides present. For example, 5% Fe oxide is required to make the clay red (pers. comm. Dr Fathi Shaqour, 10/5/2016). The plant material used for tempering the samaga is referred to as tibn which is chopped cereal straw, a by-product of the crop processing sequence after threshing and winnowing.
House walls were constructed out of stone and mud mortar and plastered over inside the house with layers of tibn tempered samaga (Figure 2(a-c)). Mud bricks were sometimes used to construct internal walls (Figure 2(c,d)). In addition to its use as wall plaster, the samaga and tibn mix was important in the construction of household 'furniture', such as shelves, niches and sleeping platforms, as well as on floors and to shape hearth features. It was also used to make the cloche-like dome of the local type of bread oven, a tabun, still in use in Jordan to the current day (Ebeling and Rogel 2016).
The house roofs were built using juniper beams (Juniperus phoenicea L.) as a support, with carefully aligned reed stems (Phragmites australis (Cav.) Trin. ex Steud.) on top (Figure 3(a)). This was followed by a shrubby packing layer of bilan, or Thorny Burnet (Sarcopoterium spinosum (L.) Spach.), a low, dense, profusely branching spiny sub-shrub from the Rosaeceae family (Figure 3(b)). The bilan is topped by a thick layer of sediment with stony inclusions ( Figure  3(c)), with a top layer of the tibn tempered samaga plaster for water-proofing, which was regularly reapplied before winter.
Many of the internal features found at Al Ma'tan are reminiscent of those found in southwest Asian Neolithic sites. For example, the food storage bins are made of clay and are similar in construction to those found at Neolithic sites such as Çatalhöyük ( Figure 4); the niches carved into the walls of the buildings are similar to those found at Neolithic WF16 ( Figure 5); and the hearths set into the floors are frequently found in many southwest Asian Neolithic sites ( Figure 6). Furthermore, the subsistence strategies at Al Ma'tan were comparable to those practiced by Neolithic populations with the villagers keeping sheep and goats and practising small-scale agriculture.
The proximity of the original occupants to the abandoned village, and their enthusiasm to help us, meant that we were able to talk at length to the former  occupants of the buildings we sampled in order to understand how the buildings had been constructed and used. Ethnographic information was obtained verbally through interviews conducted by Palmer. Information was obtained regarding construction, the use of space and the life use history of the settlement, with specific information being recorded for the buildings and areas sampled. Off-site locations of clay sources used for building construction were also sampled with the help of former inhabitants who took us directly to the clay sources used in the construction of the buildings we sampled.

Aims
The aims of this study are three-fold: (1) to determine if certain activity areas (e.g. middens, hearths and floor areas) and construction materials (e.g. roofing materials, plasters and mortars) in a southwest Asian modern traditionally built village have unique phytolith and geochemical signatures that can be used to identify these same areas archaeologically; (2) to assess how effectively portable x-ray fluorescence (pXRF) scanning can be used in the field to determine geochemical signatures from a range of past activities; and (3) to determine if the two forms of evidence, phytoliths and geoechemsitry, are more powerful for interpreting activity areas and construction materials when statistically analysed together or in isolation.

Survey and Excavation
Fieldwork at Al Ma'tan was conducted in April 2014. During this time a full topographical survey of the village was conducted using a Total Station (Leica TC407) and all buildings were planned at a 1:20 scale ( Figure 7). Three well preserved houses were selected for analysis, Buildings 1 (B1), 10 (B10) and 65 (B65) (Figures 8-10), along with a range of external areas and features. Samples were classified based upon coherent groups that best represent 'activities' or 'construction makeup/types' (Table 1).
In order to analyse the targeted locations some minor excavation was carried out to remove collapsed structural material and mud in-wash in Buildings 10 and 65. In addition, four small trenches were excavated to a depth of approximately 45 cm in Building 1 to allow the original building floors to be analysed and sampled. This was necessary because, post-abandonment, this building had been used for animal penning and, as a result, dung had accumulated.

Ethnographic Data Collection
The former inhabitants of Al Ma'tan were informally interviewed during the 10-day fieldwork season in 2014 and during subsequent visits to the village. Palmer is fluent in colloquial Arabic and undertook the interviewing with assistance from Firas Bqa'in (then a staff member for the Council for British Research in the Levant), Hussein Shabatat (a former resident of the village) and Emad Drous (Ministry of Tourism and Antiquities). Some of the informants were questioned about the history of the village and its development from the initial construction of the primary building (Building 1). Other informants provided information directly related to specific houses which we had selected for detailed analysis. Our selection of the three houses (Buildings 1, 10 and 65) was based on their exceptional preservation and the availability of former residents to provide us with the necessary information for our research; the former inhabitants of these three houses supplied accurate information to categorise the samples and to compare against the scientific analyses. Many of the older informants had memories of living in the village, collecting raw materials, constructing the houses and undertaking regular maintenance of them. Former residents provided details about construction practices, agricultural practices, herding practices, use of fuel, duration of occupation, family sizes and time of abandonment. All the information collected provided a robust background for understanding the village itself and more specifically the contexts that were being sampled and analysed from Al Ma'tan for this study.

Portable X-Ray Fluorescence (pXRF)
The locations selected for investigation at Al Ma'tan were analysed geochemically in situ (where possible) using a Niton Xl3t GOLDD+ pXRF analyser in the mining mode with the addition of helium purging from an attached portable helium canister. pXRF is a versatile and rapid technique that lends itself to a wide variety of sample types and a full suite of thirtysix elements can be recorded during analysis in the field. There are numerous highlighted limitations, such as low precision, variable analytical accuracy, lack of instrument calibration and data correction. Also ignorance of the effects of surface morphology and difficulties with quantification due to a lack of appropriate standards can all negatively impact results  ( Goodale et al. 2012;Frahm 2013). When compared with laboratory-based XRF, the results have been found to be comparable for some elements (e.g. zirconium and strontium), but not for others (e.g. barium) (Goodale et al. 2012, 882). In comparison to other methods for the determination of elemental composition such as inductively coupled plasma mass spectrometry (ICP-MS), x-ray fluorescence and pXRF do not produce as accurate results. However, while ICP-MS is a versatile technique that can achieve limits of detection (LODs) many orders of magnitude lower than XRF, samples must be in liquid form, which often requires acid digestion and laborious sample preparations. While precision and accuracy using handheld analysers are not as great as with benchtop instruments (Craig et al. 2007;Frahm 2013;Piercey and Devine 2014), pXRF offers researchers the opportunity to gain readings in the field, alleviating the need for sample export which is becoming increasingly problematic in areas of southwest Asia, and can also be used as a prospection tool to select specific samples which, if promising, can be chosen for analysis using benchtop XRF or ICP-MS if more precise and accurate readings are needed.
Preliminary field tests were carried out to establish analytical timings in order to obtain low errors  n/a n/a n/a n/a n/a 2 2 Control type 2 n/a 1 n/a n/a n/a 5 6 Control type 3 n/a n/a n/a n/a n/a n/a 0 External/Courtyard 4 0 2 n/a n/a n/a 6 Midden n/a n/a n/a 14 n/a n/a 14 Animal occupation 4 0 0 n/a n/a n/a 4 External fire installations and ashy deposits n/a n/a n/a n/a 3 n/a 3 Internal fire installations and ashy deposits 4 3 2 n/a n/a n/a 9 Hearth-make-up 3 5 1 n/a n/a n/a 9 Floors and surfaces 8 9 18 n/a n/a n/a 35 Human occupation/accumulation 0 0 0 n/a n/a n/a 0 Plasters and clay features 6 11 12 n/a n/a n/a 29 Storage features 0 7 1 n/a n/a n/a 8 Platforms and benches 5 1 1 n/a n/a n/a 7 Mortars 0 2 1 n/a n/a n/a 3 Roofs and roofing materials 3 3 3 n/a n/a n/a 9 (<10%). The pXRF analyser ran for a total of 210 s (main filter-60 s, low filter-40 s, high filter-20 s and light filter-90 s). The light filter needed a prolonged period of time for analysis to successfully detect the lighter elements such as Mg, silicon (Si) and P. The results of the analyses were examined for both elevations and depletions in elemental values, and for comparable patterns of major element concentrations. It was not our intention to use pXRF as a 'like for like' alternative to a benchtop XRF but to use pXRF as a portable instrument with a full appreciation of the limitations of this technique when compared to benchtop alternatives.

Sampling for Phytolith Analysis
Bulk sediment samples of approximately 50 g were taken for phytolith analysis from the same locations as the geochemical scanning using a clean trowel. Sample locations were photographed, added to plans and recorded using the Total Station (Leica TC407). The samples were double bagged and exported to the Department of Archaeology, Anthropology and Forensic Science, Bournemouth University for laboratory processing and analysis. In total, 144 samples were taken from a range of activity areas and construction materials (Table 1).

Modern Plant Reference Collection
A modern plant reference collection of 68 different plant species was processed to create a phytolith reference collection for the identification of phytolith morphotypes. The modern plant samples were collected in 2013 from a variety of locations in Jordan (see Vos 2016 for further details).
Terms for each morphotype were allocated using the International Code for Phytolith Nomenclature (Madella, Alexandre, and Ball 2005). A minimum of 250 single-celled phytoliths and up to 50 conjoined or multi-celled phytoliths (where possible) were identified, counted and recorded. Weight percent of phytoliths for each sample was calculated based on the original weight of sediment processed (phytolith weight % = weight of phytoliths/weight of plant matter processed * 100).

Data Reduction
Thirty four elemental concentrations were determined from the pXRF scanning procedure. A considerable number of these elemental components had a high proportion of results under the limits of detection (< LOD) and individual values with high (> 10%) error readings (two-sigma precision value) which were removed from the analyses. However, a few low detection variables were retained because they are indicative of anthropogenic activities (following Reimann, Filzmoser, and Garrett 2002), for example elements such as P, chlorine (Cl) and Mg. This resulted in 12 elements being retained for further statistical analyses. For the remaining elements that had infrequent < LOD values recorded, numerical alterations were made to deal with the missing data. All < LOD values were replaced with their corresponding lower limit of detection which is the value provided as an error reading by the analyser (as recommended by the instrument's manufacturers-Niton).
Correlation Analysis (CA) was subsequently used to assess how the remaining 12 geochemical variables covaried and to establish if there were any significant relationships between any two or more variables. This was important because in multivariate statistics correlated variables can bias the analysis. CA was conducted in a free downloadable statistical program called PAST (Hammer, Harper, and Ryan 2001) using standardised (mean of 0, standard deviation of 1) data. Both Pearson's Product Correlations coefficients (after transformation, see below) and Spearman's Rank Correlation coefficients (before transformation) were used as a measure of the degree of correlation. The strength of the correlation was judged using the published r-value. Any correlations which were higher or lower than +0.75 and −0.75 respectively, and had a p-value of below 0.05 in both correlation tests were deemed strong. For elements found to have strong correlations, one of the correlated elements was retained whilst the other was removed from further analysis to avoid using two very similar variables. Linked variables included: 1) Al and Si); 2) Si, titanium (Ti) and Al; 3) Ti and Fe; 4) zirconium (Zr) and Ti; and 5) Ca and strontium (Sr). As a result, we excluded Al, Si, Fe, Zr and Sr from the analysis and retained Mg, K, Ca, P, Ti, sulphur (S) and Cl.
Only single cell phytolith percent data were used in this study. In total, 39 single cell phytolith variables were identified and the phytolith counts were converted into percentage data for each phytolith type and sample. Variables containing less than 0.1% of the total phytolith count were removed from further analysis, leaving a total of 28 phytolith types.

Data Exploration
The selected elemental concentration data and the phytolith percentages were checked prior to statistical analyses to assess whether the observed values were sensible and to exclude any erroneous data. Through visual inspection, no analytical errors were identified but some data outliers were noticed. These data outliers were not errors but investigative outliers that resulted from anthropogenic sources, and as such were retained because they contained useful information about human practices.
The combined dataset was characterised and further inspected both visually and statistically for normality and skewness. Each variable had a different distribution shape, were often positively skewed and consisted of closed sum data (compositional data that are parts of a whole). Normalisation procedures were thus applied to the dataset to help it approach a normal distribution, reduce the compositional closure and lessen the effect of the data outliers; making most inferential statistics more applicable. Various monotonic transformations were trialled for the geochemical data, and while most made individual elements approximate normal, they did not work for the dataset as a whole. A centred log ratio (clr) transformation (Aitchison 1986) was found to be the most appropriate transformation for allowing all the geochemical variables to approximate or reach normality. Clr-transformations were conducted on the elemental data using a statistical application called CoDaPack v2 (Comas-Cufí and Thió-Henestrosa 2011). For the phytolith results, an arcsine square root transformation was made in Microsoft Excel to the percentage data. This choice of transformation moves very high or low values towards the centre, reducing the impact of common phytolith types on rarer ones. It also controls for the non-normality of the percentage data which is constrained to 0-100% and can contain many low percentage values (McDonald 2014).
Principal Component Analyses (PCA) were performed in PAST (Hammer, Harper, and Ryan 2001) using the corrected and scaled data. PCA were run on a correlation matrix for the geochemical, phytolith and combined (geochemical and phytolith) datasets. Whilst transforming values prior to statistical analysis did make the data approximate normal and brought the variances more in line, there was still some difference in the variances. Therefore, it was necessary to standardise the variables first (PCA on a correlation matrix) so that they all had variance 1 and mean 0 to enable components to be found that best represented the variation in the original data, without being overly biased by those variables that showed the most variance. Standardisation also minimised the influence of different scales and units, and improved the significance of rarer phytolith types. A PCA ordination based on a correlation matrix showed differences in the variable compositions, rather than the variable concentrations as with a variance-covariance matrix. For the phytolith and combined PCAs, it was also deemed appropriate to ensure that the variances between groups was maximised to ensure that associations between variables and data groupings were made visually clearer.
PCA analysis was conducted on the geochemical and phytolith datasets individually (Figure 11(a,b)), and in combination (Figure 11(c)) to explore the internal data structure, to detect multi-variable Figure 11. PCA biplots of the first two principal components (PCA1 vs. PCA2) for (a) the geochemical data; (b) the phytolith data; (c) the geochemical and phytolith data combined. PCA was conducted on normalised values using a correlation matrix i.e. each variable makes the same contribution to the analysis. Coloured dots represent individual samples assigned to a category type. Coloured addition symbols (+) mark the centroid of samples for each category. Black text delineates the variables driving most of the variance within the dataset, with the length of each dashed line defining the strength of each factor (longer lines = driving more variance). associations, look for similarities between observations and visually interpret results. In this section, the geochemistry results will be discussed first, followed by the phytolith results, and finally the geochemical and phytolith results combined.

Geochemistry Results
The geochemical variability of the Al Ma'tan samples can be seen in 34 elements. Of the elements measured by the pXRF, Mg, Si, K, Ca, P, Fe, Ti, Al, Sr, S, Cl and Zr form the bases of our interpretations. High values were recorded for Ca (45273-293558 ppm) and Si (15084-189929 ppm), while values for Mg, P and Sr were much lower. Small value ranges were found for Ti (4029 ppm), Sr (538 ppm) and Zr (300 ppm). Sample values for K, Fe and Al are relatively high but some smaller values also exist. Sulphur is distinctive in that while its mean value is relatively high, its median value is much lower because of a few high values inflating the mean. Unmeasurable concentrations of Mg (23%), P (65%), S (6%) and Cl (13%) were recorded but these elements were kept in the analyses because concentration levels were of importance to the study and contained information about specific anthropogenic features of interest, such as hearths.
Samples show a clear patterning of element depletions and additions depending on the assigned category type (Figure 12). For example, control/background samples are relatively high in Mg, Fe, Ti and Al, whereas most other categories show depletions in these elements. Animal occupation and midden samples have high values of K and Cl, and relatively high P. Elevated Ca, and closely associated peaks in Sr, are documented for the following categories: plasters and clay features; storage features; floors and surfaces; platforms and benches; and roofs and roofing material. In contrast, lower concentration levels of Ca are recorded for animal occupation and mud mortar samples. Uniquely, roof and roofing material samples show elevated Zr and external fire installations and ashy deposit samples show elevated Si. The high average S value in the control/background samples is biased by the S concentration of just one clay-rich sample (control type 1). Higher S values are also found in internal fire installations and ashy deposit samples and in plasters and clay feature samples.
The samples studied revealed a number of interesting points if the differences are considered by category type. First, there is a significant depletion in some elemental values in comparison to the control samples. Second, the geochemical patterns for certain categories can be similar. Third, a marked elevation in K, P and Cl is apparent for specific categories and finally, high S concentrations are often associated with specific samples and not categories.
The first two axes for the geochemical PCA account for 68.3% of the overall variance in the dataset. Only one principal component was seen as statistically significant using the broken stick method (Frontier 1976;Legendre and Legendre 1998) but with bootstrapped data then axes two and three can also be considered significant. The first two axes have been selected for graphical illustration only because they represent a high proportion of the overall differences, and are significant or close to significant in both unbootstrapped and bootstrapped analyses.
The resulting PCA (Figure 11(a)) shows welldefined groupings of samples which relate closely with the category assigned to each sample. Driving PCA axis 1 (PCA1) is the distinction between higher levels of Ti, Ca and Mg at the positive end and higher levels of Cl, S and P at the negative end. Driving PCA axis 2 (PCA2) is the difference in levels of P and K. In this regard, the PCA identifies two main latent variables or gradients that differentiate anthropogenic from natural sources on the one hand (PCA1) and the presence of animals and/or burning on the other (PCA2). Elements that associate with natural sediment sources are therefore Mg and Ti [and previously (see 'Methods' section) excluded correlated elements such a Fe and Al], whereas elements which are attributed to anthropogenic sources include S, Cl and P. These loadings imply that PCA1 defines a geochemical gradient that is driven by the presence or absence of lithogenic sediments (e.g. clays, colluvium), and their reduced signal due to more anthropogenic additions into the sediment matrix. The loadings also suggest a strong gradient associated with the presence of animal dung and burnt sediments that define PCA2; attributed by elevated levels of P and K respectively.
There is a clear discrimination of the category types according to the geochemical PCA results ( Figure 11). Mortar, roofs and roofing material, and hearth makeup all show chemical concentrations similar to the control/background samples and are very distinct from animal occupation samples. The location of these samples at the positive end of PCA1 suggests they all have a common lithogenic signature when compared to animal occupation samples which have very little lithogenic influence. Other similarities are found for storage feature, floors and surfaces, platforms and benches, and plasters and clay features categories. These samples plot towards the centre of PCA1 and the negative end of PCA2, and are aligned closely to elevated S and Ca. The broadest chemical variation can be seen with internal fire installations and ashy deposits. External fire installations and ashy deposits and internal fire installations and ashy deposits plot beside each other and likely reflect similarities in higher K and P levels, elements typically associated with burnt ashy deposits (see Custer et al. 1986 Midden samples also plot closely with the two burning categories, indicating the comparable importance of high K and P in Al Ma'tan middening contexts; probably because the middens principally consist of dumped tabun oven ash (pers. comm., former Al Ma'tan residents).

Phytolith Results
Of the 144 samples analysed, 97% of the phytolith assemblages were dominated by monocotyledonous (monocot) phytoliths ( Figure 13). A total of 57% of the samples had less than 2% dicotyledonous (dicot) phytoliths and 8% of the samples consisted of between 10 and 85% dicot phytoliths ( Table 2). The samples with the highest percentage of dicots were samples from the hearth make-up and control/background categories (specifically control type 1) ( Table 2). Categories dominated by samples with low percentages of dicot phytoliths were: plasters and clay features; middens; and platforms and benches ( Table 2).
The majority of the samples in the categories analysed are dominated by common grass morphotypes; for example: elongate dendritic, elongate smooth, elongate sinuate and rondel short-celled phytoliths. Furthermore, there are some categories with limited variability in the phytolith assemblages beyond these forms, such as: external fire installations and ashy deposits; middens; plasters and clay features; and platforms and benches.
One category which has a wider variety of phytolith morphotypes is floors and surfaces. In addition to the common grass phytolith morphotypes outlined above, there are low but significant numbers of papillae, platey, globular smooth, short celled bilobes and block phytoliths. Other category differences to highlight are the low, but consistent, presence of dicot morphotypes in the hearth-make-up and internal fire installations and ashy deposits categories; sheet and block phytoliths were identified in the hearth-makeup samples, and platey and sheet phytoliths in the internal fire installations and ashy deposits samples.
Reeds (Phragmites) can be identified as conjoined forms (Metcalfe 1960;Ollendorf, Mulholland, and Jr 1988) and further inferred from the identification of single-celled keystone bulliform phytoliths ( Figure 13). Single-celled keystone phytoliths were identified in 104 of the samples analysed and conjoined reed forms in 89 of the samples. Categories where reed phytoliths where plentiful were: roofs and associated deposits; internal fire installations and ashy deposits; external fire installations and ashy deposits; and middenswhile plasters and clay features were often devoid of reed phytoliths.
Conjoined husk phytoliths which were identifiable to genus in the samples included Triticum (wheat -57 samples), Hordeum (barley -83 samples), Avena (oat -2 samples) and Lolium (ryegrass -8 samples) ( Figure 13). Conversely, conjoined husk phytoliths that were identifiable to genus level were low in the hearth make-up and mortars categories with an average of <1.5% of the total conjoined phytolith assemblages. Conjoined husk phytoliths identified to genus level were high in the following categories: external fire installations and ashy deposits; middens and external/ courtyard with an average of >10% of the total conjoined phytolith assemblages (12.5-20.6%). The highest average percentage of identifiable husk conjoined forms was from external fire installations and ashy deposits with 20.6%. Conjoined husk phytoliths identified to genus level comprised between 4.9 and 9.1% in the majority of the remaining categories: animal occupation, floors and surfaces; internal fire installations and ashy deposits; plasters and clay features; platforms and benches; roofs and roofing materials; and storage features. The only category where all the samples analysed had conjoined forms from husks was storage features. All identifiable husks from storage features were identified as Triticum and Hordeum.
The PCA biplot for the phytolith results ( Figure 11  (b)) comprises the first two PCA axes only, which account for 52.1% of the total variance observed within the dataset. Using the broken stick method, axes 1-3 were deemed significant for the phytoliths which in total account for 66.3% of the total variance. Phytolith types that are commonly associated with grasses, and in particular grass inflorescences, including elongate smooth, rondel, elongate dendriforms, papillae and cork cells, all produce positive loadings on PCA1. At the opposite end, phytoliths that are typically attributed to dicots (e.g. sheet, block and globular smooth) as well as hair, keystone and stomata forms produce negative loadings. This distinction most likely separates sediments that contain cereal plant parts with those higher in wood/shrubs and reeds. At the positive end of PCA2, the variances seen are predominantly driven by hair base and saddle phytoliths, but also polyhedrol plain, papillae and globular smooth types. At the negative end, elongate echinate, elongate sinuate, elongate trapeziform, rondel and bulliform phytoliths dominate. Block type phytoliths also drive some of the distinction for negative PCA2.
Distinct category clusters in the phytolith data are less visually apparent; probably because of the greater variability in the dataset compared to the geochemical data and the larger number of variables included within the analysis. Reasonably well-defined comparable phytolith signatures for each category can however be seen for the storage features, platforms and benches, and  (20), Midden (9), Platforms and benches (6), Storage features (5), Roofs and roofing materials (4), External/courtyard (4), Control type 2 (3), Internal fires installations and ashy deposits (3), Animal occupation (2), External fire installations and ashy deposits (2), Hearth-make-up (1) Moderate 2-9 50 35 Floors and surfaces (13), Plasters and clay features (6), Midden (5), Roofs and roofing materials (4), Internal fire installations and ashy deposits (4), Hearth-make-up (4), Storage features (3), Control type 2 (3), Animal occupation (2), Mortars (2), Platforms and benches (1), Control type 1 (1), External fire installations and ashy deposits (1), External/courtyard (1) High 10-24 8 6 Internal fire installations and ashy deposits (2), Hearth-make-up (2) Floors and surfaces (2), External/courtyard (1), Mortar (1) Very high 50-85 3 2 Hearth-make-up, Control type 1 plasters and clay features categories which are strongly associated with elongate forms due to their location at the negative end of PCA2. Other comparable categories include internal fire installations and ashy deposits, hearth make-up, and mortars which are influenced by the presence of dicot material and plot at the negative end of PCA1. Midden and external fire and ashy deposit samples cluster at the positive end of PCA1 with cereal phytolith types and are further influenced by the presence of hair bases. Floor and surface samples are relatively widespread suggesting a diverse phytolith signature for samples attributed to this category. Whilst the centroid for floors and surfaces plots at the centre of both PCA axes, and close to elongates and rondels which are both monocot forms, floor and surface samples can also be seen plotting with dicot signatures (negative PCA1) and cereal signatures (positive PCA2). Therefore, it could be argued that on-floor activities can be more of a driver of the phytolith signature compared to the floor matrix signature in specific cases because of the mixed geochemical signals recorded.
External/courtyard samples are also distributed at opposite ends of the biplot with some samples aligning close to the influence of hair base phytoliths (positive PCA2) and others with elongate phytolith types (negative PCA2). Animal occupation samples likewise show a lack of sample clustering. There is further dissimilarity in the control/background samples so it is hard to compare other category signatures against the background observations. Their location at the negative end of PCA1 and positive PCA2, however, suggests associations with globular smooth, and reed phytolith types (e.g. keystone bulliform and saddle), more than other categories.

Combined Geochemistry and Phytolith Results
The combined PCA which includes both the phytolith and geochemical data (Figure 11(c)) is most comparable to the phytolith PCA. In total, thirteen axes were created during the combined data PCA but only the first four axes were deemed statistically significant. The first four axes have combined eigenvalues explaining 74.9%. Due to the difficulty of understanding four latent variables only the first two will be discussed here.
Along PCA1 we see that Ti and Ca co-vary with hair, globular smooth, sheet, block, bulliform, keystone bulliform and stomata types at the negative end. For positive PCA1, Cl and P play an important role in delineating samples plotted here, as do elongate dendriform phytoliths and papillae phytoliths. Cl and Ca are key geochemical drivers of negative PCA2, as are the leaf/stem phytolith types of rondels, elongate sinuate and elongate trapeziform, and block phytoliths. Mg drives a lot of the variance seen for positive PCA2, as do saddle and hair base phytoliths.
The biplot (Figure 11(c)) of the first two principal components for the combined dataset shows some separation of clustered category samples (see Table 3 for summary of clusters). The first noticeable cluster (C1) includes control/background, mortar, and hearth make-up categories which are distinct because of higher levels of Ti, Ca and Mg, and phytoliths akin to woody plant and reed species. The second group (C2) includes storage feature, plasters and clay features, platforms and benches, and floor and surfaces categories which plot together because of elevated Ca and Cl, as well as leaf/stem phytoliths of elongate form and rondels. The third group (C3) includes animal occupation, external fire installations and ashy deposits, and midden categories, and variables such as P, S, grass inflorescence phytolith types and hair bases. External/courtyard samples are too varied to cluster but samples plot within each of the three clusters (C1-3) identified above and consequently signatures must relate more to specific activities conducted within the external setting rather than the make-up of the external setting. Internal fire installations and ashy deposits are again too diverse to provide a distinct group but they do not plot with the second cluster suggesting a lack of elongate phytolith types for samples of this origin. The centroid for internal fire installations and ashy deposits is central on both PCA axes and whilst this implies a mixed signal for these observations, it also suggests some relation to higher levels of K, which is an element commonly associated with burnt wood ash and would be expected to be inflated in fire-related deposits (see Holliday 2004; and references therein).

Geochemistry
Lithogenic Influences: Mortars, Hearth-make-up, and Roofs and Roofing Materials The geochemical results from Al Ma'tan showed clear divisions between activity contexts with categories such as mortars, hearth-make-up, and roofs and roofing materials consisting of mainly lithogenic elemental concentrations with other categories such as animal occupation having little lithogenic influence. We know from the ethnographic evidence that the mud mortars used in the wall construction are comprised of unaltered sediments obtained from areas close to the village. These were then mixed with water and used between the stones as an adhesive. The category hearth make-up is comprised of the sediment directly beneath the hearth, not the remnant fuel, and therefore consists of sediments or clays that are naturally deposited beneath the building prior to construction. Similarly, a large portion of the roofing material derives from unaltered colluvial sediments collected from the surrounding landscape which was placed on top of the beams, reeds and shrubby material as shown in Figure 3.
Anthropogenic Influences: Animal Occupation, Storage Features, Floors and Surfaces, Platforms and Benches, and Plasters and Clay Features Categories such as animal occupation have little lithogenic influence and represent a more anthropogenic input. Typically anthropogenic influences were recognised in contexts such as: animal occupation; storage features; floors and surfaces; platforms and benches; and plasters and clay features (see the section 'Data Exploration').The animal occupation deposits have a reduced lithogenic sediment signature because they consist primarily of animal-faecal material and plant remains, and comprise virtually no local sediments. Contexts related to burning, internal fire installations and ashy deposits, external fire installations and ashy deposits and middens, also exhibited clear patterns (see the section 'Data Exploration').
The contexts which are relatively high in Ca and S (storage features, floors and surfaces, platforms and benches, and plasters and clay features), geochemically reflect the selection of samaga 'clay' used in plaster construction and the addition of materials such as gypsum. Internal fire installations and ashy deposits, external fire installations and ashy deposits and middens have similar signatures due to the elevation of two elements which commonly result from exposure to fire -P and K (Holliday 2004). Middens can be seen as indirect indicators of activities and represent the end product of those activities (Shillito 2017). The midden deposits sampled at Al Ma'tan predominantly consist of ashes from the tabun ovens which the local inhabitants informed us were fuelled with dung and are located next to the midden.

Phytoliths
The distinction between the categories from the phytolith results is less clear in comparison with the geochemical results. However, there are some divisions between categories which are supported by the ethnographic evidence.
Addition of Temper: Storage Features, Platforms and Benches, and Plasters and Clay Features Storage features, platforms and benches, and plasters and clay features plot together in the PCA analysis (Figure 11(c)) and these categories are influenced by elongate phytolith forms. These are from the plant material which was used as temper which the former inhabitants informed us was preferably barley tibn because it was softer than wheat tibn. Conjoined phytoliths positively identified as barley were identified in approximately half of the samples in these categories.
Dicotyledonous Signatures: Internal Fire Installations and Ashy Deposits, Hearth-makeup, and Mortars Other categories which have similar phytolith signatures are internal fire installations and ashy deposits, hearth-make-up, and mortars. These categories are influenced mainly by dicot forms with a paucity of conjoined monocot forms which could be identified to genus level. From the ethnographic information these categories are known to be made from natural clay sources which are littered with shrubs, and that tibn is not used in their construction. Therefore, the phytolith results match the ethnographic information gathered about these categories with the dicot material being naturally derived from the local clay source. The final category influenced by dicots, internal fire installations and ashy deposits, has a signature which represents wood fuel selection for the internal hearths. This correlates with the information obtained during the ethnographic investigation which was that only wood fuel was used inside the buildings.
Cereals: External Fire Installations and Ashy Deposits, Midden and Storage Features Categories which are influenced by phytoliths indicative of cereals are the external fire installations and ashy deposits, and middens. As noted above, the percent of conjoined husk phytoliths which could be identified to genus level was high in these categories. These categories contain dung remains because they are comprised of the dung-fuelled tabun ovens and the adjacent ashy rake-out from them. The prevalent husks identified from these categories are Triticum and Hordeum which was supplied as fodder to the animals demonstrating that the phytolith information accurately reflects the ethnographic information for these samples.
Similarly, the only category where all the samples analysed had conjoined husk forms is storage features. All identifiable husks from storage features were identified as Triticum and Hordeum which correlates with the ethnographic information from the former inhabitants who told us that these storage features had contained cereals.

Diverse Phytolith Signatures: Floors and Surfaces
There are some categories which have a more diverse phytolith signature, for example, the floors and surfaces. This varied signature in the floor samples is indicative of the different activities carried out on the floors within the buildings with samples being from a range of areas such as sleeping areas, entrance-ways, kitchen areas, areas adjacent to hearths, areas near to storage features and peripheral areas. Furthermore, material identified on floors and surfaces in some archaeological contexts has been observed to be transported and re-deposited from other contexts in the form of 'foot traffic' on the soles of peoples feet or shoes (Shillito 2017). Spreads of ashes on floors could produce a diverse signature more similar to hearth deposits (Regev et al. 2015). Therefore, the variation in phytolith signatures within this category is expected.

Combining the Proxies
When data from both the geochemical and phytolith analyses were statistically analysed together, the grouping of categories reflects the results from the phytolith analysis because there are a greater number of distinct variables in the phytolith assemblage than is found in the geochemical data. Some categories were identified as being too varied by the combined geochemical and phytolith PCA to form clusters. These categories were external/courtyard and internal fire installations and ashy deposits. The external/courtyard samples could be different because of external influences such as wind and rain which would deplete the geochemical and phytolith signatures or alternatively could concentrate them in certain courtyard areas. Similarly, the external/courtyard areas have a varied life history ranging from human use to post occupation animal penning areas and finally complete abandonment which all play a role in the geochemical signature witnessed i.e. the more diverse the range of activities, the more diverse is the resulting phytolith and geochemical signature. The variation between samples from the category internal fire installations and ashy deposits, could result from varied cooking activities. Other differences could reflect differences in how the hearths were managed when they fell out of use.
Overall, it is apparent from our results that these two proxies are better statistically analysed separately rather than together. Results demonstrated that when trying to combine the results to run the statistical analysis the outcome was less powerful than the results considered for each proxy individually. The proxies provide different but complementary information that should then be interpreted together after statistical analysis. For example, geochemistry can help in identifying areas of burning through the presence of elevated levels of P and K, while phytolith analysis can help pinpoint what type of fuel was being used, for example dung or wood. In terms of how this relates to our first research aim it can be concluded that phytoliths and geochemistry can be effective in defining activity areas within southwest Asian Neolithic sites but are best statistically analysed separately.
Another important outcome of this research was in demonstrating that in situ pXRF analysis can be effectively used to help understand and interpret southwest Asian ethnographic contexts. Undoubtedly, more precise and accurate concentration results could have been obtained by using a laboratory-based instrument but this would have greatly increased costs both in terms of time and money (Frahm 2013;Frahm and Doonan 2013;Speakman and Shackley 2013). We found that in situ pXRF was adequate for our needs and we were able to see relative differences between certain elements within our samples both in terms of elevations and depletions from the natural control samples. This enabled us to understand the broader geochemical signatures left behind from certain activities. Based on the results from this study we would also suggest that pXRF is an effective tool for site prospection and can help identify promising contexts within the field that can be sampled and exported for further geochemical analysis if greater precision and accuracy are needed.
One potential limitation, but a key result of this research was that many of the different context types we analysed were, in fact, comprised of the same natural source material i.e. the samaga clay and tibn straw mix used to plaster walls and surfaces as well as to construct internal installations. This inevitably caused some blurring and mixing of the different categories. This was a due to the construction methods employed at Al Ma'tan but the same phenomenon can also be observed in southwest Asian Neolithic sites which are also largely built using natural resources. It may be possible to make distinctions between these categories and features in other types of archaeological sites which are built using different construction techniques. This complication is one of the key justifications for our intention to continue this research by utilising micromorphology as an additional proxy to complement and supplement the geochemical and phytolith analyses. Micromorphological analyses can distinguish different depositional pathways even if the material is similar macroscopically, geochemically and in its phytolith content. The results of the integrated micromorphological analyses will be published in a forthcoming paper.
Overall as a pilot study based on 144 samples the dual phytolith/geochemical method provided results that could help identify these context types in southwest Asian Neolithic sites and could potentially be more effectively used on sites which do not rely so heavily on the same natural resources in their construction. The next stage of this research will involve cross-comparing our results with the phytolith and geochemical results from the sites of Wadi Faynan 16 (Pre Pottery Neolithic A) and 'Ain Ghazal (Pre Pottery Neolithic B) to verify the efficacy of these methods (Figure 1). In addition, micromorphological sampling will be conducted on blocks from Al Ma'tan to further our understanding of contexts on a microscopic scale.

Conclusion
The results of the analysis of the plant phytoliths and geochemistry from the ethnographic village of Al Ma'tan demonstrates that both forms of evidence are valuable in providing information about activity areas and building practices in traditionally constructed southwest Asian villages. The most important finding from this study is that when used in combination these proxies can provide vital insights into activity patterns and construction practices. However, the results also show that when running the statistical analysis the two proxies are best analysed separately because the greater number of variables in the phytolith assemblage dilutes the results from the geochemical analysis. The results provide different, but complementary information which can then be interpreted together to collect as much information about the analysed contexts and provide the fullest picture of the patterns left behind. When the two forms of evidence are used in combination their interpretative power is greater than when they are used individually. This is because many of the context categories and building materials were comprised of the same local clays and had a high organic matter content both of which lead to an elevation in the same chemical elements (Wilson, Davidson, and Cresser 2008;Milek and Roberts 2013). In these instances phytoliths provided an additional source of information which gave clarity to the data and allowed the categories to be more fully understood.
Micromorphology samples were also collected during fieldwork from Al Ma'tan to help further our understanding of the activities and formation processes. These samples will be used to examine some of the specifics of the results presented here. The combination of geochemistry, phytolith analysis and micromorphology is a rapidly growing area (Shillito 2011;Shillito et al. 2011;Mallol et al. 2013) and is proving powerful in its interpretive ability.
The fact that many of our results tallied with the information provided by the former inhabitants of the site leant confidence to our results and demonstrated that this integrated dual-method can provide valuable information about southwest Asian sites. In the future we intend to test this method on Neolithic sites from this region but we also suggest that this method is not restrictive either geographically or temporally but is a global method that can be used in any time period.
Elliott, S. Is a Post-Doctoral Research Fellow at the British Institute in Amman (Council for British Research in the Levant). She is a Geoarchaeologist specialising in phytolith analysis and micromorphology and her research focuses on early Neolithic sites in the Middle East and Geoethnoarchaeology. Palmer, C is Director of the British Institute in Amman, the Council for British Research in the Levant's institute in Jordan. Carol's research interests lie in the anthropology and archaeology of food production and consumption, particularly in the rural Middle East, and rural transformations in response to the social, developmental and political upheavals of the modern period.
Grattan, J is Acting Vice Chancellor at Aberystwyth University with a background in environmental and archaeological science. John has worked extensively in the Wadi Faynan region of Jordan and has a research focus' on geochemistry, the emergence of metallurgy, and volcanism.