5.1 A ¹⁴C chronology

Seven ¹⁴C dates on procellariid bone from sinkhole sites 1710-1 and 9659-1 (Table 5) provide the data needed to construct an absolute chronology for the stratigraphic sequences of Kalaeloa sinkholes (¹⁴C age determinations are listed fully here because, to the best of our knowledge, all previous listings are either incomplete or contain errors; the information in Table 5 is from records kept at Beta Analytic dating laboratory). Davis designed the dating project carefully, selecting bone samples for dating from each of three depositional units present at the two sites. Because of this, the dating calibration can integrate relative stratigraphic information by use of a Bayesian statistical framework (Buck et al. 1996; 1992; 1991). In this framework, information on relative ages of dated events is used to constrain the calibrated ages of dated samples; the calibrated age of a sample will always be younger than the calibrated age of a sample recovered from a stratigraphically older deposit, regardless of the relative ¹⁴C ages of the two samples. Thus, samples that yield inverted ¹⁴C ages are restored to their correct relative ages, as this relationship is defined stratigraphically. Typically, addition of stratigraphic information to the calibration procedure improves the ability to interpret the results archaeologically. In addition, adoption of a Bayesian framework provides a way to obtain age estimates for events that were not directly dated, which is useful in this case because it is possible to estimate ages of depositional unit boundaries. Under the assumption that changes in depositional modes were penecontemporaneous across the region, age estimates for depositional unit boundaries derived from sites 1710-1 and 9659-1 can be extrapolated to depositional sequences of sinkhole sites that were not dated. A primary objective of the Bayesian ¹⁴C calibration reported here is to estimate calendar ages of transitions from one depositional unit to the next.

The analysis requires three assumptions: 1) sediment deposition in sinkholes is continuous - there are no hiatuses between or within depositional units; 2) there is no significant hiatus between the death of the bird whose bone was dated and deposition of the bone in the sinkhole; and 3) post-depositional movement of dated bones through the stratigraphic column was not sufficient to change their positions relative to depositional unit boundaries. We believe that these assumptions are reasonable at this stage of analysis, but do not believe that they are universally valid. In our view, further study of the dynamics of sediment deposition in sinkholes is clearly warranted.

Given the stratigraphic and ¹⁴C information, and assumptions listed above, it is possible to formulate a model of the relationships among depositional units and unknown calendar ages of events represented by seven ¹⁴C dates. Following standard practice, we indicate the lower boundary of depositional unit i (i = I, II, III) as β_i and the upper boundary as α_i. Let θ_j denote the unknown calendar date BP of event j (j = 1..7). Then archaeological and ¹⁴C information from the two Kalaeloa sinkholes can be expressed in the form of the following inequalities.

β_III > θ₇ > α_III = β_II > θ ₆ > θ ₅ > θ ₄ > α_II (1)
β_II > θ ₃ > θ ₂ > α_II = β_I > θ ₁ > α_I (2)

This model was implemented using the OxCal software package (Ramsey 1995). Seven ¹⁴C determinations associated with the θ_i (Table 5) were calibrated with a marine curve (Stuiver and Braziunas 1993) using a δr value of 110±80 established for ocean waters surrounding the Hawai'ian Islands (Dye 1994b).

Estimates of the calendar ages of the dated events are listed in Table 6 as 2 σ highest posterior density regions, along with calendar ages reported by Davis (1990) and Christensen (1995). The estimates yielded by Bayesian analysis are younger by 200-700 years than age estimates reported by Christensen (1995), as expected. They are, however, very close to the results yielded by Davis' calibration procedure, with two important exceptions. The two exceptions are θ₇, at the early end of the sequence, and θ₁ at the late end of the sequence. The Bayesian estimate for the age of θ₇ is 400-600 years earlier than Davis' estimate, and the estimate for θ₁ is 200-400 years later. These differences have the effect of doubling the estimated duration of the interval between the earliest and latest events in the sequence, transforming the 400-1,000 year sequence posited by Davis, to one that spans 1,000-2,000 years.

Figure 3: Estimated ages of depositional unit boundaries. Left, boundary of basal diagenetic and structural collapse deposits; right, boundary of structural collapse and transported sediment deposits. Solid lines above the x-axis indicate 67% and 95.4% highest posterior density regions

The 95.4% highest posterior density region yielded by Bayesian calibration for the estimated age of the boundary of basal diagenetic and structural collapse deposits is 50 BC-AD 950 (Fig. 3), an interval that spans current estimates of the date of initial Polynesian colonisation of the islands. The date of colonisation has become a point of contention, over which roughly two schools of thought have formed. There is an argument for 'early' colonisation dating to the AD 100-400 range (e.g. Kirch 1985; Hunt and Holsen 1991) and an argument for a 'late' colonisation, as late as AD 600-1000 or even AD 800-900 (e.g. Spriggs and Anderson 1993; Athens et al. 1997). Events in the basal diagenetic deposit, exemplified by θ₇, are likely to have occurred either very early in the Polynesian era, or before Polynesian colonisation of the islands. Although the early colonisation estimate is coeval with the boundary of basal diagenetic and structural collapse deposits, it is unlikely that Polynesians would have settled or farmed this marginal region soon after colonization (Tuggle 1997) and it is safe to say that events in basal diagenetic deposits pre-date significant Polynesian activities at Kalaeloa. It is not possible with the data at hand to estimate with confidence when basal diagenetic sediments were first deposited. This event pre-dates event θ₇, however, and a reasonable inference is that deposition of basal diagenetic deposits began more than 2,000 years ago.

The 95.4% highest posterior density region for the estimated age of the boundary of structural collapse and transported sediment deposits is AD 1320-1740, an interval that encompasses the last four centuries of the pre-Contact era (Fig. 3). The structural collapse deposits represent at least the first half of the Polynesian era, but given uncertainties in the date of Polynesian colonisation and in the age estimate of the depositional unit boundary, these deposits might encompass nearly the whole of the era. Events in transported sediment deposits either took place late in the pre-Contact era, or in the period after Contact. The bone dated for event θ₁, whose estimated age falls late in the Polynesian period, was collected from the bottom half of the transported sediment deposit at site 9659-1 and is consistent with this assessment.

© Internet Archaeology URL: http://intarch.ac.uk/journal/issue10/dye/kalaeloasu1.html
Last updated: Tue May 29 2001