Elsevier

Journal of Archaeological Science

Volume 52, December 2014, Pages 633-644
Journal of Archaeological Science

Sites forlorn: dating intervals of abandonment at three shell middens on Santa Cruz Island, California using Bayesian chronological models

https://doi.org/10.1016/j.jas.2014.03.029Get rights and content

Highlights

  • I establish Bayesian chronological models for three shell middens on Santa Cruz Island.

  • This approach facilitates formal age estimates for two distinct intervals of abandonment.

  • The earlier more protracted interval of abandonment lasts from ca. 2629–1698 cal. BP.

  • This interval is shown to correspond to significant cultural and environmental shifts.

  • I highlight implementation and benefit of a Bayesian approach to building site-chronologies.

Abstract

Archaeologists interested in variable mobility patterns and the factors that influence them rely on chronometric data to illuminate repetitive cycles of occupation, abandonment, and reoccupation. This article establishes and interrogates Bayesian chronological models in order to provide formal age estimates for two distinct intervals of site abandonment at three substantial multi-component shell middens from Santa Cruz Island, California. The earlier, more protracted interval of abandonment (ca. 2629–1698 cal. BP) following terminal Early Period occupation at each site location represents a temporal datum for a series of significant cultural, demographic, and technological developments throughout the region. I propose that this interval of localized site abandonments indicates a period of reorganization provoked by a shift in the paleoenvironment, a preliminary hypothesis that requires further testing. This research calls for development of similarly rigorous site chronologies on Santa Cruz Island and throughout the region that will shed further light on the validity of the temporal gaps documented in this analysis and most certainly will facilitate the development of additional hypotheses. More broadly, this article highlights the implementation and benefit of a Bayesian approach to building and interrogating site chronologies.

Introduction

Mobility is a crucial component of human adaptations to environments. Climate, population size, and technology—among other cultural and economic factors—structure movement to and between resource locations. Archaeologists seek to unravel this dense web of interactions by examining accretions of cultural material that accumulates sequentially over time (Stein et al., 2003:297). Consideration of variable mobility patterns, and the factors that influence them, requires chronological precision capable of establishing the true order of events within a given context, and discerning whether events in different contexts are actually contemporaneous (Culleton et al., 2012:1573). Thus, rigorous chronological control is central to inferences about mobility strategies. Repetitive cycles of occupation, abandonment, and reoccupation of favored locales, complicate this pursuit. Stratigraphically complex sites, such as shell middens, are notoriously difficult to decipher (Stein et al., 2003:298). However, several major methodological, analytical, and statistical improvements for building archaeological site chronologies have coalesced during the last decade (see Bayliss, 2009, Bronk Ramsey, 2008a, Bronk Ramsey, 2008b, Buck, 2004). In this paper, I draw upon these advances to establish and interrogate high-precision chronologies for three multi-component shell middens from Santa Cruz Island, California. This refined chronological framework illuminates intervals of site abandonment and facilitates comparison with detailed paleoenvironmental and subsistence reconstructions to shed light on Late Holocene mobility patterns on the Northern California Channel Islands.

The rationale for a Bayesian approach to constructing site chronologies is not only to arrive at more realistic age estimates for individual measurements (sensu Bayliss and Whittle, 2007, Bicho et al., 2012, Culleton et al., 2012, Jazwa et al., 2013, Kennett et al., 2011, Nakamura et al., 2013, Outram et al., 2010, and many others) but also to provide a statistical framework for further analysis (Bronk Ramsey, 1998:463,472). Although the individual dates are of considerable significance, archaeologists are most interested in the temporal parameters of cultural processes and human behavior inferred from chronometric data (Nicholls and Jones, 2001:503). As soon as conventional radiocarbon dates are received from the lab, archaeologists begin the process of interpreting these data. Inevitably, this involves a series of assumptions, some as simple as to whether a series of dates represents multiple dates of a single synchronous event or whether those same dates represent multiple sequential events. An experienced archaeologist will judge the former or the latter to be more plausible based on prior information, such as stratigraphic relationships between contexts containing dated samples (see Bronk Ramsey, 1995:426; 2008a:264). Use of these priors informs interpretation, regardless of whether they are made explicit.

A Bayesian approach to chronology building forces archaeologists to make explicit assumptions about the priors, and provides a quantitative framework to formalize and test assumptions. Agreement indices for the individual calibrations and for the model as a whole quantify the correspondence between the prior and posterior distributions, while accounting for the correlation between parameters (Bronk Ramsey, 2000:201; 2009:357). This statistical measure allows archaeologists to evaluate the validity of assumptions, such as the assumed stratigraphic integrity of a site, on which the model is constructed. A second statistic, convergence, measures the degree to which there is a truly representative and stable solution to the model (for more information, see Markov Chain Monte Carlo samplers, Bronk Ramsey, 1995:429; 2001:359; 2009:353). These two statistics allow archaeologists to evaluate both the reliability and stability of constructed chronological models. Moreover, a well-constructed model can produce reliable date estimates (boundaries) for events not directly dated by single radiocarbon determinations, such as the beginning and end of deposition (Bayliss et al., 2007:6), and can be queried to derive useful information about the temporal duration of an activity (span) or hiatuses between events (intervals). “Interrogating the model”, as Bronk Ramsey (1998:463) refers to such queries, affords archaeologists the opportunity to more effectively extract and evaluate information regarding the significance of temporal patterns. Herein lies the true value of constructing Bayesian site chronologies—the opportunity to move beyond visual inspection of groups of calibrated dates (Bayliss et al., 2007:8–9; Whittle and Bayliss, 2007:21).

In this paper, I begin with a Bayesian approach to integrate non-quantitative contextual information (the priors), with probability distributions from radiocarbon dates (the standardized likelihood) to trim confidence intervals and refine age estimates (the posteriors) for three depositional sequences. However, I focus analysis and discussion on the temporal information derived from querying the final chronological models.

Section snippets

Background

The Late Holocene represents a period of demographic, cultural, and economic development for the Chumash of the Santa Barbara Channel Region of California. Significant population increase (Glassow, 1999:56; Kennett et al., 2009:310) parallels the development of new maritime technologies (Arnold, 2007, Gamble, 2002, Rick et al., 2002, Rick et al., 2004), craft specialization (Arnold, 1992, Arnold, 2001, Arnold, 2004), and the emergence of social hierarchies (Winterhalder et al., 2010). Most of

Materials and methods

The explicit purpose of my excavation at CA-SCRI-236, CA-SCRI-823, and CA-SCRI-568 was to collect samples appropriate for detailed chronological, paleoethnobotanical, zooarchaeological, and isotopic analyses. These samples provide the foundation for evaluating diachronic variation in mobility and associated patterns of resource use. My principal concern was to ensure comparability among the assemblages recovered from each site. At least one volumetrically controlled unit was excavated by trowel

Results

In the resulting chronological models, the posterior modeled date ranges (displayed in black) are trimmed slightly in comparison to the standard calibrations (displayed as an outline). The site sequence-based models (Fig. 2, Fig. 3, Fig. 4, Fig. 5) provide estimates for the beginning and end of each period of site occupation; whereas the period phase-based model Fig. 6 provides general estimates for the beginning and end of occupation associated with each temporal period. The agreement index

Discussion

The protracted hiatus in site occupation that occurred between the terminal Early Period and Middle Period (ca. 2629–1698 cal. BP) at CA-SCRI-236, CA-SCRI-823, and CA-SCRI-568 coincides with recent recognition of a 1500 year gap in the occupation of interior site locations across Santa Cruz Island ca. 2750 BP-1250 cal. BP (Perry and Glassow, 2014:16)3

Conclusion

Bayesian chronological models highlight and provide formal age estimates for two distinct intervals of localized abandonment at three substantial habitation sites located in the western sector of Santa Cruz Island. The earlier, more protracted hiatus following terminal Early Period occupation at each site location represents a temporal datum for a series of significant cultural, demographic, and technological developments throughout the region. If the chronological sequences identified at these

Acknowledgments

This article stems from a paper presented at the SAA's in 2013 in a Symposium organized by Robert L. Kelly and Nicholas Naudinot. The final manuscript has benefitted greatly from comments and criticism provided by Michael Glassow, Barbara Voorhies, Danielle Kurin, Anne Pisor, and two anonymous reviewers. Douglas Kennett and Brendan Culleton provided not only the variable local reservoir correction factors but have also provided a sounding board throughout the development of these chronological

References (55)

  • J.E. Arnold

    Complex hunter–gatherer–fishers of Prehistoric California: chiefs, specialists, and maritime adaptations of the channel islands

    Am. Antiq.

    (1992)
  • J.E. Arnold

    The Origins of a Pacific Coast Chiefdom: the Chumash of the Channel Islands

    (2001)
  • J.E. Arnold

    Foundations of Chumash Complexity

    (2004)
  • J.E. Arnold

    Credit where credit is due: the history of the Chumash oceangoing plank canoe

    Am. Antiq.

    (2007)
  • A. Bayliss

    Rolling out revolution: using radiocarbon dating in archaeology

    Radiocarbon

    (2009)
  • Histories of thee dead: building chronologies for five southern British long barrows [special issue]

    Camb. Archaeol. J.

    (2007)
  • A. Bayliss et al.

    Bradshaw and Bayes: towards a timetable for the Neolithic

    Camb. Archaeol. J.

    (2007)
  • J.A. Bennyhoff et al.

    Shell bead ornament exchange networks between California and the Western Great Basin

    Anthropol. Pap. Am. Mus. Nat. Hist.ory

    (1987)
  • N. Bicho et al.

    Chronology of the Mesolithic occupation of the Muge Valley, Central Portugal: the case of Cabeço Da Amoreira

    Quat. Int.

    (2012)
  • C. Bronk Ramsey

    Radiocarbon calibration and analysis of stratigraphy: the Oxcal program

    Radiocarbon

    (1995)
  • C. Bronk Ramsey

    Probability and dating

    Radiocarbon

    (1998)
  • C. Bronk Ramsey

    Methods, materials, and instruments – comment on “The use of Bayesian statistics for 14C dates of chronologically ordered samples: a critical analysis”

    Radiocarbon

    (2000)
  • C. Bronk Ramsey

    Development of the radiocarbon calibration program

    Radiocarbon

    (2001)
  • C. Bronk Ramsey

    Deposition models for chronological records

    Quat. Sci. Rev.

    (2008)
  • C. Bronk Ramsey

    Radiocarbon dating: revolutions in understanding

    Archaeometry

    (2008)
  • C. Bronk Ramsey

    Bayesian analysis of radiocarbon dates

    Radiocarbon

    (2009)
  • C.E. Buck

    Bayesian chronological data interpretation: where now?

  • Cited by (8)

    • Testing the Malthusian model: Population and storage at Housepit 54, Bridge River, British Columbia

      2018, Journal of Archaeological Science: Reports
      Citation Excerpt :

      It is possible that the charcoal used for this sample could have originated in stratum IIa1. We did not engage in further modeling of boundaries and phases in OxCal (e.g. De Souza et al., 2016; Kennett et al., 2014; Overholtzer, 2015; Thakar, 2014) given the short accepted distribution for IIa-IIo, the stratigraphic integrity of the floor and roof deposits, and the degree of inter-floor variability within the overall range. With an estimated 346 year range for the formation of the IIa-IIo floor sequence, we can project that on average each floor was occupied for approximately 23 years, close to ethnographic expectations for standard use life of a traditional Mid-Fraser roof superstructure (Alexander, 2000).

    • Beyond culture history: Coast Salish settlement patterning and demography in the Fraser Valley, BC

      2016, Journal of Anthropological Archaeology
      Citation Excerpt :

      Comparing the rise and fall in number of large settlements to the size and location of these settlements (e.g., Schaepe, 2009) would add to our understanding of these thresholds and to community organization. An important area of inquiry is understanding the social and ecological contexts that encourage people to stay put, to split into smaller groups, or to aggregate into larger groups (cf. Angelbeck, 2009; Angelbeck and Cameron, 2014; Angelbeck and Grier, 2012; Hakar, 2014; Hickok, 2013). As noted by other researchers, there is some correspondence between the periods in which people chose to aggregate into large settlements and other culture historical trends in the region.

    • The timing of postglacial coastal adaptations in Eastern Iberia: A Bayesian chronological model for the El Collado shell midden (Oliva, Valencia, Spain)

      2016, Quaternary International
      Citation Excerpt :

      To validate the model robustness, the convergence C values should be higher than 95%. Finally, Oxcal is equipped with a set of modelling tools such as the Outlier() command, which allows it to identify those problematic dates that violate the stratigraphic order excluding them from the chronological model; or the Outlier analysis, an averaging method which assumes the same prior probability of being outliers for all the radiocarbon dates considered in a given model to produce down weighted posterior probability distributions (Bronk Ramsey, 2009a) All of these analytical capabilities and indices make Oxcal a suitable tool for building robust chronologies on shell midden sites with complex depositional histories (Kennett and Culleton, 2010; Bicho et al., 2013; Thakar, 2014; Wicks et al., 2014). For this work, we have produced a three-phase Bayesian sequential model of El Collado chrono-stratigraphy using 16 AMS radiocarbon determinations.

    • Bayesian <sup>14</sup>C analysis, formation processes, and accumulation rates of the Prisoners Harbor shell midden and village complex, Santa Cruz Island, California

      2015, Journal of Archaeological Science: Reports
      Citation Excerpt :

      All 14C dates were adjusted using the Intcal13 or Marine13 datasets for terrestrial and marine samples, respectively (see Reimer et al., 2013). Marine samples were further adjusted using a ∆R of 261 ± 21 based on dating of Olivella and Mytilus (see Jazwa et al., 2012; Thakar, 2014). Researchers working on California's Channel Islands have used a variety of different ∆R values, including 225 ± 35 and 261 ± 21 (see Kennett et al., 1997; Jazwa et al., 2012).

    View all citing articles on Scopus
    View full text