«By Nathan B. Goodale A dissertation submitted in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY WASHINGTON STATE ...»
Summary Through the examination of the prehistory of the southern Levant from the Early Epipaleolithic to the end of the Pre-Pottery Neolithic period, we witness many fascinating trends in the evolutionary history of human societies. These include the appearance of the first sedentary communities; the development of storage technologies; subsistence intensification, which lead to the domestication of plants and animals; the rise of the first town-scale communities; and the ability to potentially support craft artisans. The rich archaeological record of the area, coupled with the abundant data (as well as consistently recorded variables of proxies of human
particularly interesting case study in which a paleodemographic model of the substantial increase in human numbers associated with the NDT can be developed. In the next chapter I cover the methodology utilized to accomplish this task.
An important aspect of building an accurate but versatile demographic model from the archaeological record is selecting pertinent and viable data. Foremost, a model must incorporate multiple variables appropriate for population reconstruction.
Second, the data must be readily available from archaeological literature from several geographic areas. Importantly, each proxy should be a standard measure that researchers commonly present in site reports as well as in larger discussion pieces regarding changing settlement practices and mobility strategies.
The purpose of this chapter is twofold, first, to outline the proxies of population growth used to interpret population growth rates during the transition to agriculture in the southern Levant, and second, to introduce new analytical techniques of assessing population demography using multiple variables representing population proxies. Importantly, variable proposed has been utilized as a measure of human population density by archaeologists and paleodemographers and is readily available in archaeological literature. The measures presented here include 1) habitation area defined in hectares, 2) depth of deposits in meters, 3) frequency of sites occupied in
occurring in 50-year increments where each date is counted in every 50-year interval that its 95 percent confidence interval spans. In other words, if a date has a calibrated
increments. As specified below, certain criteria were established to determine if specific data could be used based on its integrity in terms of both accuracy (if the date makes sense to the original excavator) and precision (the error range fit within some expectation).
The second section of this chapter covers the analytical technique crosscorrelation (or convolution analysis) and its potential utility to directly analyze and compare these variables as proxies of population. The cross-correlation analysis technique is used to examine how well each variable correlates to the others with the hypothesis that if all of these variables are monitoring and tracking the same phenomenon (population growth), then there should be a strong cross-correlation.
Proxies of Population Growth In this section the proxies utilized in the paleodemographic models demonstrating past population growth are outlined. Each variable is utilized as a proxy of human occupation or in other words, higher values should be reflective of higher numbers of people.
enabled archaeologists to examine large-scale spatiotemporal problems (Chamberlain 14 2006). C date distribution has been regularly utilized as a population proxy as a
Bocquet-Appel and Demars 2000; Brantingham et al. 2004; Fort et al. 2004; Gamble et al. 2004; Goodale et al. 2004, 2008a; Housley et al. 1997; Pettitt 1999; Rick 1987).
Rick (1987) argued that radiocarbon date frequency can be representative of human population densities, and their distribution can reflect changing settlement systems or population replacements. 14C dates have been utilized to trace the demise of the last Neanderthals in Europe (Bocquet-Appel and Demars 2000; Pettitt 1999), paleodemography and the transition to resource intensification in the Pacific Northwest of North America (Goodale et al. 2004, 2008a), continental colonization of Europe after the last glacial maximum (Fort et al. 2004; Gamble et al. 2004; Housley et al. 1997) and the Americas (Anderson and Faught 2000) as well as detecting the hiatus of human occupation during the middle Holocene in the Pampas of Argentina (Barrientos and Perez 2005). In all of these studies both the frequency and the distribution of 14C dates is an important parameter in the signal of human activity and population densities.
While temporal frequency distributions have been a common method in order to measure the extent of human activity, the analytical technique has not gone without criticism (Surovell and Brantingham 2007). The basic premise of the critique is that in most instances the frequency distributions of radiocarbon dates as well as sites take the form of a positive curvilinear function (Surovell and Brantingham 2007). In the traditional context, this would indicate that frequency increases as human activity increased; however, this is also the exact outcome if one assumes that taphonomic
farther one goes back in time, the weaker the demographic signal (Surovell and Brantingham 2007).
Briefly, in attempt to explore issues of taphonomic bias (Surovell and Brantingham 2007:1874) in these frequency datasets, I first compare the value for 1/λ (the value to control the rate of drop off in the decaying exponential model) for all of the variables. Second, while Surovell and Brantingham (2007) are correct that the systematic destruction of sites through time should be modeled as a decaying exponential, they do not consider the rate at which sites are added. In order to compensate for this, the final section of Chapter Seven presents a computer simulation taking into account both the rate of site destruction and additions to the archaeological record. The results of both of these analyses suggest that the data utilized to build the population growth rate model are not severely impacted by taphonomic bias. Additionally, as suggested by Surovell and Brantingham (2007:1874) to correct for taphonomic bias, short-term variation (50-year increments in this study) is superimposed over long-term trends (22-8kya). This technique is appropriate because while taphonomic bias will influence the long-term trend, it
confidence interval. It is expected that if 14C dates correlate with population increase, there should be a statistically significant positive correlation through time, even without the affect of taphonomic bias.
14 C Dataset Most of the 14C dates utilized in this study can be found in the C14 radiocarbon
original references where those dates were published from the Near East, Africa and Europe. While this resource is available, it did not provide the dataset utilized in this study. Instead, it served as a reference point, as all data compiled to build the demographic model in this study were pulled from the original literature. While compiling the data for the model, my relationship with the CONTEXT dataset was somewhat of a double-edged-sword. In many instances I found typographical errors in the CONTEXT database, missing references, and data I found that was not included. However, the database did aid in making sure that I could trace back as much data as possible to build the demographic model and make it as precise,
criterion for a 14C date to be included in the analysis is that it could be traced back to the original literature. If a date could not be found in the original literature or through personal communication with the principal investigator, the date was not included in
questioned the relationship between the timing of the date in association with the archaeological materials, the date was not included in the analysis. Lastly, if the 14C date has an error range exceeding ± 10 percent of the mean of the date, the date was
interval. In total, the population of 14C dates utilized in this study is N = 520, where the range fell within the time frame of 20,000 to 8,000 cal BP from the Early Epipaleoltihic through to the end of the Pre-Pottery Neolithic.
Habitation Area in Hectares Habitation area or site size refers to the horizontal extent of archeological deposits. Site size has been considered in relationship to population predominantly to estimate how many people occupy some given amount of space based on ethnographic accounts that are then projected into the past (Adams 1965; Kramer 1978), although the actual ability to do this has been questioned (Kuijt 2000). Kuijt (2000) demonstrates that the figures for estimating how many people actually occupy a defined amount of space vary greatly cross-culturally and through time. Despite this problem, there does appear to be a fairly strong correlation between area
these disparate viewpoints, in this study I do not attempt to place any figure of actual human numbers occupying a defined unit of space and instead only rely on site size as a relative proxy of human population size. The use of the data in this manner is also appropriate for examining long-term trends in population growth and decline, where the densities of humans within a defined area most likely changed dramatically. Thus, one figure of human numbers per defined area is insufficient for this study.
Habitation area is probably the most common variable presented for prehistoric archaeological sites in the southern Levantine literature. Predominantly, habitation area is calculated for the Epipaleolithic in square meters and in hectares for the Neolithic. Site size is normally estimated by the spatial extent of archaeological materials. All data utilized here have been converted into hectares. Habitation size was counted for each 50-year increment that the 14C data indicated the site could have been occupied indicated by the 95 percent confidence interval. In this analysis, the full site size was assigned all of the time periods. In the future, I would probably do this differently by proportioning the amount of site size to the likely hood ratio that the calibrated date fall. Nonetheless, it is expected that if site size is a relatively accurate predictor of increasing populations, there will be a statistically significant positive correlation with time.
Depth of deposit or site volume has been utilized, although rarely, as a proxy of population density. Ammerman et al. (1976) directly attribute the depth of deposits in connection with other lines of evidence including the amount of time that the site was occupied, and the volume of interior house materials in order to estimate the number of houses occupied at any given time. With an assumption of five people per house, Ammerman et al. (1976) then calculate how many people may have occupied the site at a time.
Similar to the issues discussed above concerning habitation area, in this study I do not superimpose any specific numbers of humans in relation to the depth of deposits. Instead, depth of deposits is used as a relative proxy of human population where deeper deposits represent more intense human occupation and thus, a greater number of humans. Similar to site size, depths of deposits were monitored for each 50-year interval that the site could have been occupied indicated by the 14C 95 percent confidence interval. It is expected that if depth of deposits is a relatively accurate predictor of increasing populations, there will be a statistically significant positive correlation with time.
Frequency of Sites Occupied The frequency of sites occupied at any give time directly relates to the larger regional settlement patterns reflecting the larger use of the landscape (Adams et al.
2001; Waters and Kuehn 1996). While the frequency of sites occupied through time
even a negative correlation through time may be indicative of population packing events rather than actual increasing population numbers. This would be indicative of increasing population density and humans packing into fewer aggregates. In other words, a signature of population packing should be a non-statistically significant positive correlation or a negative correlation. In opposition, a statistically significant positive correlation would be indicative of not only increasing population but also increasing population density. This is supposes that depth of deposits, site size, and 14 the frequency of C are statistically positive, likely signifying increased population growth through time. In this study, the frequency of sites occupied was graphed by the associated 14C date(s) 95 percent confidence interval for every associated 50-year increment.
includes simple linear regression, a parametric regression analysis technique utilized to predict the outcome of one variable from another. Since regression analysis assumes normality, the raw data values for each variable were transformed into
regression analyses were conducted with the computer statistical program SAS v.8.
The other analytical technique utilized in Chapter Seven is cross-correlation analysis.
As this technique has not been used to my knowledge with archaeological datasets, a
The cross-correlation analysis was conducted with the mathematical computation program, Mathematica v. 7.0.