The Importance of Multiple 14C Dates from Significant Archaeological Contexts
- First Online:
- Cite this article as:
- Levine, A. & Stanish, C. J Archaeol Method Theory (2014) 21: 824. doi:10.1007/s10816-013-9177-4
- 509 Downloads
Radiocarbon (14C) dates are the most important means for determining the age of Holocene archaeological deposits. The theoretical physical basis of this method is by now unassailable, having been consistently tested and refined over two generations. However, the means by which this method has been applied and the interpretation of these results remain as key issues, particularly for complex archaeological discoveries that substantially affect our understanding of world prehistory and social evolution. Many factors can produce uncertainty or variation in the 14C concentrations of samples, even those that have been selected from the same archaeological context or event. A number of recent studies have also addressed the ways in which ambiguities and irregularities in the 14C calibration curve can affect the interpretation of archaeological dates and temporal patterns. Of greatest concern, however, is a growing practice of using only one or two samples to date a significant prehistoric context or event. The date of these events, usually relative to other human activities, often holds important theoretical implications for evolutionary anthropology and related disciplines. In this article, we demonstrate that such a practice is rarely adequate or acceptable. Rather, proper procedure requires a suite of dates that permit statistical verification that the deposit or event itself is being correctly dated. We present a detailed case study that highlights the importance of analyzing multiple samples of 14C from significant archaeological contexts.
KeywordsAMS dating Radiocarbon calibration South America Lake Titicaca Basin Taraco
More than 50 years ago, a model for the production, distribution, and decay of the radioactive carbon isotope 14C led to the proposal that 14C content could be used to determine the calendar age of organic specimens (Taylor 1978, 1997). Early experiments by Willard F. Libby (1952, 1955) verified this method, and subsequent work addressed the techniques and demonstrated the variety and extent of its applications (Geyh and Schleicher 1990; Taylor 2000a, b; Taylor et al.1992). The development of accelerator mass spectrometry (AMS) in the late 1970s was a major revolution in 14C dating that radically increased counting sensitivity in comparison with conventional decay counting. AMS dating, which directly measures the mass of different isotopes, allowed for orders of magnitude reductions in sample size and counting times. From the perspective of counting statistics, it also allowed greater precision to be assigned to milligram size samples on a routine basis (Taylor 1996). Having established dates for such monumental transitions in human history as early plant domestication, the peopling of the New World, and the European Paleolithic, in addition to hundreds of other significant events too numerous to be listed here, the importance of AMS dating for archaeology cannot be overstated.
With AMS technology, it is possible to measure the 14C concentration of any given sample with a level of accuracy and precision that was difficult to achieve with conventional decay counting. However, many factors can affect the measured 14C concentration of samples, and each of these sources of variability can, in some cases, introduce added degrees of uncertainty to a date that can, in turn, affect interpretation. Well-known natural sources of error include variability in the initial 14C concentrations of various carbon reservoirs. For example, because of marine reservoir effects, marine organisms—and those that feed on them—appear artificially old, with 14C concentrations that do not reflect when the organism was last alive (Mangerud 1972). The hard-water effect, the result of dissolved carbonates depleted of 14C, can likewise produce errors on the scale of hundreds of years for aquatic organisms (Brennan and Quade 1997; Riggs 1984); the magnitude of these errors can vary even within a region (Moore et al.1998). A similar effect is found in areas with a history of volcanism, where deep fissures permit magmatic CO2 to percolate from the mantle to the surface (Cook et al.2001; Geyh 2000). Isotopic fractionation, the result of natural biochemical processes, can lead to variation in carbon isotope ratios that must be normalized using a measured 13C/12C ratio (δ13C value) before an accurate calendar age based on the residual 14C concentration measured in a sample can be obtained (O’Leary 1981; Olsson and Osadebe 1974; Taylor 1987).
Other sources of variation in 14C dates can be caused by actions occurring not only during sample collecting and processing but also during laboratory analysis (Kra 1986). While contamination is often discussed in terms of sample deposition or collection in the field, it also can be an issue in the laboratory. In laboratory settings, the need of most AMS systems to convert samples to graphitic carbon is a documented source of trace amounts of modern contamination (Taylor 1996). It has also been established that differences in laboratory sample preparation and analysis procedures can result in slightly different 14C ages, even for the same sample. Controlling for inter-laboratory variation can be difficult because no lab is ever consistently older or younger than any other (Hertelendi 1990; Martin and Johnson 1995). These inherent errors in 14C dating can be significant, but are controllable with good laboratory procedures that characterize university-based research facilities.
A second category of error builds off these technical issues but, in our view, involves a more serious problem that can seriously affect the validity of archaeological interpretations. That is, while the 14C-based age of the material tested can be highly accurate within the interval defined by its associated error term, what those 14C dates actually reference in the past may be, in many instances, problematic. Dean (1978) has discussed this issue extensively, differentiating among “dated events,” “target events,” and “bridging events,” which can potentially yield divergent 14C ages. It is possible that a single archaeological event will in fact yield multiple, accurate 14C dates that may not necessarily overlap at ±2σ. Here, we define an archaeological “event” as a single episode of deposition; these usually consist of a suite of features and/or cultural contexts among which a relationship has been determined based on stratigraphy and other contextual evidence. The “old wood” and “old shell” problems both refer to discrepancies between the age of an organism and when the material from that organism was last used by humans, and are classic examples of within-event variation that have been the subject of much discussion (Rick et al.2005; Schiffer 1986; Windes and Ford 1996). Variation within a material class has received less attention among archaeologists, yet is still recognized as a source of error in 14C dating (Culleton et al.2006; Keith et al.1964; Olsson 1992). However, in spite of these problems, researchers often rely on only one or two dates for significant events. We argue that this practice is inherently problematic, as the probability of obtaining the “actual” age of an event through this practice is very slim, especially in contexts where the sample materials are uncertain (i.e., “carbonized material”) or in cases of depositional contamination.
We also argue that this problem can be overcome if there are sufficient samples of carbon from known materials that will yield statistically significant results. For instance, if the archaeological context is ideal—a relatively non-disturbed, well-stratified, and sufficiently deep deposit—then a number of dates from many of the strata should yield a series of results from oldest to youngest with marginal overlap in their ±2σ distributions. It is not necessary to date each level or event multiple times if the ordering of the dates matches the relative depositional history of the sample. An example of this kind of context includes work by Levy et al., who analyzed 23 radiocarbon dates from a single stratigraphic sequence (Levy et al.2008). Their application of Bayesian statistics further enhanced the accuracy of these dates. In this case, the tight correlation between stratigraphy and the obtained dates, plus statistical treatment of the data, allows for high levels of accuracy and precision (Ahlstrom and Smiley 1998) throughout the entire dated sequence.
Multiple sites that contain evidence of the same event, allowing cross-dating over multiple, independent depositional contexts, can likewise overcome these dating challenges. An example of this situation would be Formative Oaxaca, where archaeological signatures of discrete ritual processes and state conquest are found in different sites throughout the valley. Marcus and Flannery (2004) and Spencer and Redmond (2001) were able to accurately and precisely date these events using multiple samples drawn from several sites. In this example, the ages of these similar events could be verified with through the analysis of 14C from several independent contexts.
However, most archaeological deposits and contexts are not so ideal. Most archaeological excavation contexts are characterized by numerous perturbations to their depositional history including rebuilding episodes, human and animal disturbances of the cultural levels during and after use, site abandonment and reuse leaving sterile levels between cultural ones, differential deposition rates, flooding, and aeolian deposition, as well as a host of other anthropogenic and taphonomic factors. In these instances, one or two dates from a single significant level or event are insufficient to date the context or feature. Instead, the accurate and precise dating of these contexts requires the analysis of multiple radiocarbon samples.
We use the following case study from the Lake Titicaca Basin of Peru to show that multiple dates are necessary when dealing with significant archaeological features from these less-than-ideal contexts. The reasons for this are threefold. First, with multiple samples it is possible to test for statistical consistency among the dates. Second, certain inconsistencies among dates from a single archaeological event may be resolved through their aggregation, which allows for the determination of a pooled mean calendar age. Third, due to inherent variation among materials, it is critical to know what is being dated, such as grasses, seeds, logs, trees, leaves, food stuffs, and so forth because these have different use-lives in any social context. Likewise, it is important to know how that material was deposited—either by human activity or by natural processes, such as aeolian, colluvial, or alluvial deposition—in order to control for carbonate contamination, as well as other cultural and taphonomic biases.
Radiocarbon Dating and the Titicaca Basin Formative
The Formative Period of the northern Titicaca Basin represents the emergence of the first complex societies in the region. The Formative is conventionally divided into three parts—Early, Middle, and Upper—each of which is associated with major socio-political transformations. The early Formative, beginning ca. 2000 bc, corresponds with the first permanent villages in the region and the earliest widespread use of pottery. The Middle Formative, beginning around 1400 bc, is associated with the first evidence of ranked societies and the emergence of a settlement hierarchy in the region (Plourde 2006; Stanish 2003). During this time, some villages became regional centers characterized by semi-subterranean stone enclosures and sunken courts. By the beginning of the Upper Formative, ca. 500 bc–ad 100, a number of competing centers with multiple sunken court complexes had developed. Of these numerous centers, Pukara emerged as the single dominant polity in the north Basin by no later than ad 100.
As in many areas of the world, understanding the development of complex societies in the Titicaca area relies heavily on radiocarbon dating for determining the ages of archaeological deposits. Calibration of these radiocarbon dates, which compares 14C measurements with calendar ages provided by other independent dating methods, such as dendrochronology or U-Th ages from corals (Bard 1998; Bard et al.1990; Reimer et al.2009), is necessary to correct for variation in atmospheric 14C over the past 40,000 years, and to determine a precise calendar age range for archaeological samples. Bard (1998) discusses three of the most significant sources of variation in the 14C/12C ratio. Centennial scale variations can be “linked to cosmic-ray modulation by the magnetic properties of the solar wind” (Bard 1998:p. 2036). A gradually increasing intensity of the Earth’s magnetic dipole has caused fluctuations over the last 30,000 years. Internal changes in the carbon cycle have affected atmospheric ∆14C. Furthermore, due to the instability of the 14C/12C ratio over time, the calibration curve does not follow a straight line, and while each calendar year theoretically corresponds with a radiocarbon year, the reverse is not always true (Stuiver and Suess 1966).
Substantial periods of the Titicaca Basin Formative are also affected by irregularities in the radiocarbon calibration curve, specifically by a significant plateau occurring 2540–2400 bp (Levine 2012; Whitehead 2007), as well as a number of other smaller plateaus, reversals, and cliffs (steep drop-offs). Plateaus in the calibration curve are sections in which the slope is significantly less than 1.0, and correspond with periods of relatively constant 14C fluctuation (Blackwell et al.2006). Radiocarbon ages that intersect with such plateaus can yield artificially long—and therefore non-precise—age ranges, as well as asymmetric or discontinuous calendar age ranges. “Wiggles” in the curve, formally known as de Vries effect variations, represent short-term oscillations in atmospheric 14C levels, and tend to have the same effects as plateaus on calibrated dates (Bamforth and Grund 2012; Manning 2006–2007). Steep areas of the curve produce the opposite effect, compressing relatively wide radiocarbon year intervals into much smaller calendar age ranges. These types of problems associated with ambiguities and irregularities in the 14C calibration curve compound long recognized issues related to sampling, contamination, and variation in laboratory procedures that can produce variation in 14C concentrations.
Dating a Major Event at Taraco
During the early part of the Upper Formative, Taraco was one of two sites competing for regional dominance in the northern Basin. The site is located along the Ramis River approximately 15 km north of Lake Titicaca. The significance of the Taraco area has long been noted by scholars impressed by the quantity and quality of elaborately carved stelae found in and around the eponymous modern town (Chávez and Chávez 1975; Kidder II 1943; Lumbreras and Amat 1968; Mujica 1978; Neira Avendaño 1962; Rowe 1942; Tschopik 1946). The archaeological site is composed of a principal mound surrounded by a dense cluster of contemporary settlements that are linked by a network of roads and causeways. Systematic survey in the region discovered that the entire area of Formative occupation totals well over 100 ha (Stanish and Umire 2004). The site’s uninterrupted occupation from the Archaic period through modern times makes it an ideal locale to study long-term processes of social evolution in detail.
Excavations were conducted on a large artificial terrace located on the river edge just below the highest part of the mound. During the Formative Period, this area was a high-status residential sector of the site. This work revealed a stratified occupational sequence reaching 3.75 m in depth with basal levels dating to 1260–1055 cal bc (Levine 2012). The most significant level in this sequence was a major site-wide burn event in association the early Pukara phase occupation. The scale of this burn was immense, and evidence of it was found in all areas that were tested, including each of the excavation units, and in a profile cut along the margin of the river. This transect of the terrace revealed a continuous stratum of ash and architectural debris measuring at least 35 m in length that corresponded with an identical layer discovered in the excavation units. This burn event has been interpreted as an episode of deliberate destruction that effectively destroyed the high-status residential sector of the site, and marked the end of Taraco as a center of political and economic activity in the region (Stanish and Levine 2011).
Analysis of Radiocarbon from the Site-Wide Burn Event
The burn event at Taraco is extremely significant and has implications for the development of complex polities in the region and for our understanding of cultural evolutionary processes worldwide (Marcus 2008; Marcus and Flannery 2000; Spencer 1998; Stanish and Levine 2011). Our research design included three testable hypotheses relative to the date of the event: If the burn is relatively late—after ca. ad 500—then a model of competitive interaction with an emergent polity from the south, known as Tiwanaku, would be supported. If the date is early—before ca. 500 bc—then a model of internal strife would be supported since there was no other polity of sufficient complexity to compete at such a level with Taraco at that time. If the date of the burn falls somewhere in between, then a model of peer polity competition would likely be supported, the only possible rival being the Pukara polity to the northwest.
The remaining three samples were wood charcoal selected from large roof beams that were also burned in the conflagration. All of these were analyzed at the Keck Carbon Cycle AMS laboratory. Analysis indicated that these roof beams were much older than the annual grasses that had been used to thatch the roofs. Moreover, unlike the grasses, the ages of these beams were not consistent, indicating that they were harvested at different times (i.e., when they last exchanged 14C with the atmosphere). One of the beams (UCIAMS-86317) in fact pre-dates the burn event by several hundred years, dating to 750–405 bc. We argue that this date is not anomalous; rather, it is indicative of conservative behaviors typical in arid environments. These results are consistent with the expectations of the “old-wood” model, and suggest that wooden beams were not replaced with great frequency, but were instead preserved across several generations after their initial collection for use in each rebuilding of a structure.
The archaeological strata from Taraco are not ideal, a situation typical of most archaeological depositions. While we did have relatively good vertical stratigraphy given that the deepest layers were older and there was a progressively younger set of strata above, the stratigraphy also revealed a number of disturbances, including pits, intrusive walls, and reconstruction episodes within this otherwise intact depositional sequence.
The analysis of 14C from Taraco revealed distinct temporal episodes that correspond to different activity patterns in the same archeological feature. While all of the dates are physically “correct,” the dates from the annual grasses date the burn event much more accurately because they have such a short life and are not reused or recycled. They indicate that the high-status residential compound was destroyed in the first century ad, a date consistent with the expectations of a peer-polity competition model. The earlier dates from each of the beams do not reveal anything about when the compound was burned, but rather provide dates for the cutting of the beams since the beams had such long use-lives (re-used and recycled) in construction. Had we relied on just one or two charcoal fragments that derived from these beams, we would have accurately dated the time of timber harvest (when the tree died/was cut down), but would have grossly erred in determining when the house compound was burned. In turn, an incorrect hypothesis would have been supported, inaccurately reconstructing the evolutionary dynamics of the region.
These results highlight the importance of obtaining multiple dates from such significant archaeological features. This is especially evident in the six annual grass samples selected from the burn event. By combining the radiocarbon ages of the grasses from this single event, we obtained a much more precise date than had we used just one or two samples. Pooling dates of short-lived materials in this manner can be extremely useful, especially when differences among the individual dates are not insignificant. Even in cases of annual grasses—ideal materials for radiocarbon dating—the selection of any single specimen could yield a skewed or incomplete calendar age range for an event. Multiple dates are necessary to avoid such ambiguity. A very large corpus of dates can also help to identify and remove outliers or problematic dates, though that was not necessary in this particular case.
The site-wide burn event discovered at Taraco represents some of the earliest evidence for intensive organized raiding and war in the Lake Titicaca region, providing insight into the evolutionary processes at work in this area of primary state formation (Stanish and Levine 2011). The analysis of multiple 14C dates from this single event has been instrumental in reaching this interpretation. The analysis of three wood charcoal samples highlights the depositional complexity of this event, while the combined dates of six samples of annual grasses shows that the destruction of this sector of Taraco immediately predates the emergence of Pukara as the dominant polity in the region.
The concept of “old wood” has been a topic of much discussion in the American Southwest and similar desert environments. It is self-evident that such a potentially distorting factor must be controlled in the field. However, the problem for researchers is not in excluding obvious beams when dating a significant event, but rather the inclusion of small, indistinguishable fragments of burnt beams that could affect the dating of the sample and the event. Even short-lived materials will display some variability due to a number of factors. In this case, aggregating the dates from the annual grasses was necessary to obtain a meaningful and relatively precise age for this very important event. If we had only selected one (or even two) 14C samples, we could have potentially supported two substantially different evolutionary models.
While we argue that multiple dates are key to achieving accurate interpretations and reconstructions of cultural process, we realize that we must have a feasible limit on the number of 14C dates per significant context in the real world of archaeological research. To this end, we advocate a methodology that is realistic. Valid statements about the chronological position of an important context must have a sufficient number of 14C dates to be statistically significant, nothing more, and nothing less.
We thank the National Science Foundation, the Peruvian Ministry of Culture, and the Cotsen Institute of Archaeology of UCLA, T. Levy, J. Marcus, and R.E. Taylor.