The transformation of the Birnirk and Punuk cultures into the Thule culture remains central to the development of modern Inuit across the Arctic (Mason Reference Mason, Friesen and Mason2016). Although their cultural modalities shared technologies and practices, they also exhibited distinct features, ranging from aesthetic appliqué, architecture, mortuary customs, to recreation (Collins Reference Collins, Jesse and Norbeck1964; Ford Reference Ford1959). Harpoon head commonalities considered as diagnostic led Mathiassen (Reference Mathiassen1927, Reference Mathiassen1929) and Mason (Reference Mason1930) to postulate a descent relationship between the Birnirk and Thule cultures. Birnirk was also seen as “an outgrowth of Old Bering Sea” (Collins Reference Collins, Jesse and Norbeck1964:99), a preceding and partly contemporaneous cultural stage of the Northern Maritime tradition (Collins Reference Collins1960), whereas on St. Lawrence Island, Punuk is recognized as strongly sequential to the Old Bering Sea culture stratigraphically, typologically, and iconographically (Mason and Rasic Reference Mason and Rasic2019).
The chronological placement of Birnirk and Thule is imprecise despite a century of investigations (Stefansson Reference Stefansson1914), with a definition of Thule that relies on scientific excavations from the Canadian Arctic (Mathiassen Reference Mathiassen1927). The first generation of researchers lacked chronometric methods and relied on intuitive inferences from geomorphic processes. For the early Thule, Mathiassen (Reference Mathiassen1927:86) inferred an age of circa AD 1000 employing the isostatic uplift elevation of Canadian beach ridges, as compared to the established age at similar elevations in Scandinavia. Harpoon head morphological shifts led Mathiassen (Reference Mathiassen1929:54–55) to derive Birnirk from an older Thule culture, but stratigraphic data from northern Alaska established that the relationship was reversed: Birnirk was ancestral to Thule (Ford Reference Ford1959). In the 1950s, the new radiocarbon dating method placed Birnirk between AD 600 and 1000 (Ralph and Ackerman Reference Ralph and Ackerman1961), matching archaeological expectations (Giddings Reference Giddings1960). In the 1970s the Walakpa site, south of Utqiaġvik (Figure 1), provided the first dated in situ stratigraphic succession of Birnirk to Thule, even though it relied mostly on assays of seal bones influenced by the marine reservoir effect (Stanford Reference Stanford1976). The direct dating of Birnirk occurred as Morrison (Reference Morrison2001:81) analyzed five museum-archived antler harpoon heads from Piġniq, the Birnirk type site, demonstrating an overlap between late Birnirk / early Thule diagnostic types “within a generation or so on either side of AD 1000.” Mason and Bowers (Reference Mason, Bowers and Grønnow2009:29, 33) broadened the search into Kotzebue Sound, a region with strong Punuk influence (Dumond Reference Dumond2000); they were seeking the “first entry of Thule people” into Alaska, possibly circa AD 1250, as long hypothesized with the Arctic Woodland intrusion into interior Alaska (Ford Reference Ford1959; Giddings Reference Giddings1952; Mason Reference Mason2020); this dating was better aligned with the now younger chronology for the Thule eastern migration (Friesen and Arnold Reference Friesen and Charles2008; McGhee Reference McGhee, Appelt, Berglund and Gulløv2000).
Figure 1. Map of Alaska and Bering Strait larger area with names of cited archaeological sites (GIS data: Global Forest Watch 2000, Alaska Geospatial Office, Natural Earth Data).
During the last two millennia, the Bering Strait region witnessed a fluorescence of cultural expression with the interaction and competition of multiple highly mobile communities and societies (Bronshtein et al. Reference Bronshtein, Dneprovsky, Savinetsky, Friesen and Mason2016; Mason Reference Mason2020) until the Thule Inuit emerged in the last 800 years (Mason Reference Mason, Friesen and Mason2016). How and through what process—migration or the sharing of different technologies—these cultural expressions contributed to the Thule Inuit still requires systematic analyses of the artifact record (Mason Reference Mason, Friesen and Mason2016). In any event, either as part of this emergence or shortly thereafter in the thirteenth century (Friesen and Arnold Reference Friesen and Charles2008; McGhee Reference McGhee, Appelt, Berglund and Gulløv2000), the Thule Inuit expanded eastward across the North American Arctic, replacing a culturally heterogenetic landscape of Dorset and Norse communities (Friesen Reference Friesen2020). Who specifically expanded eastward into the interior and along the arctic coast (Hollinger et al. Reference Hollinger, Ousley, Utermohle, Maschner, Mason and McGhee2009; Unkel et al. Reference Unkel, Norman, Tackney, Krus, Jensen, Alix, Mason and O’Rourke2022) and precisely when and where these people acquired the cultural specificities that are used to trace their migratory expansion (Mason Reference Mason2020; Morrison Reference Morrison, Maschner, Mason and McGhee2009; Prentiss et al. Reference Prentiss, Walsh, Gjesfjeld, Denis and Foor2022) are still topics for debate. Despite the 30-year effort to increase radiocarbon dating in Alaska and the now routine use of accelerator mass spectrometry (AMS) radiocarbon assays from well-authenticated contexts, the resulting chronologies remain imprecise. Archaeologists still cannot precisely assign age ranges to the cultural components that precede Thule Inuit nor precisely establish the timing of the first Thule presence across Alaska.
The dating of the Birnirk-to-Thule transition between AD 900 and 1300 is further complicated by “de Vries” anomalies in the cosmic-ray–conditioned production of radiocarbon in the upper atmosphere (Taylor and Bar-Yosef Reference Taylor and Bar-Yosef2014:59). Consequently, even a radiocarbon determination with a narrow sigma value (e.g., 15–30 years) can calibrate within distributions of several centuries (Telford et al. Reference Telford, Heegaard and Harry2004). The resulting impasse limits our understanding of settlement patterns, human–environment relationships and interregional interactions at a critical juncture in the long cultural history of Inuit/Iñupiat. Dumond (Reference Dumond, Don and Richard2002:352) suggested a path forward: “What is needed most is closely dated series of living floors that span the past two millennia.” Only by establishing the contemporaneity of sites and of houses within sites will archaeologists precisely refine understanding of cultural units.
From the Bering Strait to the eastern Arctic, coastal inhabitants settled at key resource locations, often reoccupying the same site over several generations, thus facilitating site discovery (Mason and Gerlach Reference Mason and Craig Gerlach1995). However, houses were placed in superimposed succession of building episodes on the same footprint, producing anthropogenic mounds (Carter Reference Carter1966; Sheehan Reference Sheehan1997) and resulting in intricate “tangle of timbers” (Ford Reference Ford1959:45, 60), with younger deposits mixed or incorporated into older occupation levels. Unfortunately, nearly all the Birnirk houses or burial mounds were excavated before 1970 (Carter Reference Carter1966; Dumond Reference Dumond2000; Ford Reference Ford1959; Mason Reference Mason1930; Stanford Reference Stanford1976), with only a handful of excavations containing both Thule and Birnirk occupations. For most pre-1960 datasets, rudimentary recording procedures yielded only the barest stratigraphic context, with few absolute dates and cultural affiliation based on harpoon head typology. As a result, most occupations remain undated, as do the time frames of construction and the abandonment of houses.
In this article, we examine the timing of the Birnirk-to-Thule transition on the Alaska coast through a prism of dates subjected to Bayesian chronological modeling of two recently excavated dwellings at the Cape Espenberg site, Rising Whale (KTZ-304). We couple tree-ring ages with radiocarbon dating to constrain the timing of the house occupations. These analyses are the first in Alaska to establish the age, occupation length, and contemporaneity of two adjacent houses—one of Birnirk affinity and the other of Thule.
Study Area
The Cape Espenberg dunes on the southwest margin of Kotzebue Sound (Figure 1) attracted settlers intermittently over the last five millennia, but the commitment to sedentism did not arise until its later settlement by Birnirk and Thule people (Harritt Reference Harritt1994; Tremayne Reference Tremayne2015). The Cape Espenberg spit (Figure 2) was formed through beach ridge accretion and subsequent dune formation (Mason et al. Reference Mason, Hopkins and Plug1997), producing an exemplar of horizontal stratigraphy (Waters Reference Waters1992) known from recorded maritime hunters’ preference to settle on the beach ridge closest to the open ocean, facing the Chukchi Sea (Mason et al. Reference Mason, Hopkins and Plug1997). As new dunes grew vertically and seaward, people relocated to remain close to the sea.
Figure 2. Maps of the Cape Espenberg spit with location of beach ridges and KTZ-304 site. E-1 to E-8 are the most recent successive prograding sand dunes (Google Earth – Image ©2020 Maxar Technologies; Image ©2020 Terra Metrics © 2024 Microsoft Corporation Earthstar Geographics S10).
In accepting this settlement determinant, archaeologists assume the contemporaneity of all cultural features on a dune ridge, emphasizing only a relative sequence from older to younger (Darwent et al. Reference Darwent, Mason, Hoffecker and Darwent2013; Harritt Reference Harritt1994; Schaaf Reference Schaaf1988). Generally, the surplus of habitable land on the more than 40 dunes at Cape Espenberg encouraged new house construction across the landscape, rather than the vertical reuse that produced elevated archaeological mounds in more stable or constricted landscapes (see Dumond Reference Dumond2000; Ford Reference Ford1959). However, the surplus land assumption is not universally valid: in its midsection, the Espenberg dunes are exceedingly low, today scarcely 1 m asl, and are separated by wide marsh-filled swales. From AD 200 to 900 only two or three dunes attained an elevation over 1 m asl, and dunes were less attractive for settlement. Relevant for this study, only a single dune (E-6) had formed at the end of the spit circa AD 1000.
The theoretical interest in the Birnirk-to-Thule transition was fostered by the discovery in 2006 and subsequent excavation of a Birnirk dwelling at the Rising Whale site on E-6 dune (Figure 3), the first discovered in more than 40 years (Hoffecker and Mason Reference Hoffecker and Mason2011). The discrete nature of the houses at the site, with limited or no overlying younger cultural layers and most likely rapid post-occupation burial, reduces the probability of stratigraphic mixing.
Figure 3. Map of Rising Whale KTZ-304 site with identified houses and other features (courtesy John Darwent, University of California-Davis, 2007). (Color online)
Taphonomic Issues in Arctic Dating
Issues with contextual provenance complicate the continuing use of the “legacy” radiocarbon database of the Bering Strait region, which has a long-standing reliance on assays conducted in the 1950s–1970s that are hobbled by large error ranges (Collins Reference Collins1953; Gerlach and Mason Reference Craig and Mason1992; Morrison Reference Morrison1989; Ralph and Ackerman Reference Ralph and Ackerman1961). Early efforts to conduct radiocarbon dating of Birnirk and Thule sites in the western Arctic often focused on charcoal, wooden artifacts, or marine and terrestrial bones (Morrison Reference Morrison1989). Since the late 1970s, archaeologists have found that an uncritical use of wood or charcoal radiocarbon dates can be misleading because the “old wood” effect produces ages that are potentially considerably older than the timing of archaeological use (Dean Reference Dean and Michael1978; Schiffer Reference Schiffer1986). In the Arctic, this time offset is increased by the nature of the wood resource. In addition to local shrubs, wood available on arctic coasts is driftwood, which incorporates variable lag times between tree death, the deposition of the log, and its anthropogenic use (Alix Reference Alix, Friesen and Mason2016; Giddings Reference Giddings1942). This lag time increases with the distance between the locale of tree growth and the coast. However, transit time may be short, within a decade (Oswalt Reference Oswalt1951:8), as found for some of the dated modern drift logs from Cape Espenberg (Alix et al. Reference Alix, Juday and Grant III2012).
Despite the difficulties associated with driftwood, coastal Alaska offers the potential of determining precise calendrical ages in the tree-rings of its well-preserved wood structures, even if they are not felling dates but are that of the natural or accidental death of trees (Giddings Reference Giddings1942, Reference Giddings1952; Taïeb Reference Taïeb2023; VanStone Reference VanStone1955). After Giddings’s sudden passing in 1964 until the early 2000s, radiocarbon dating became the solution for all chronometric concerns, leading to the neglect of tree-ring dating efforts (Nash Reference Nash2000) even in northwest Alaska. Giddings’s (Reference Giddings and Theodore1962) optimism for dendrochronology is vindicated by the failure of revisionist studies (Anderson and Feathers Reference Anderson and Feathers2019; Murray et al. Reference Murray, Robertson and Ferrara2003; Shirar Reference Shirar2011) that use radiocarbon ages or luminescence-dated ceramics to better constrain the “Arctic Woodland” tree-ring chronology (Giddings Reference Giddings1952).
A reinforcing synergy arises by combining radiocarbon and tree-ring dating (Giddings Reference Giddings and Theodore1962:130): radiocarbon dates of the end-rings of well-contextualized architectural driftwood are coupled with multiple, cross-dated tree-ring sequences to establish regional dendrochronologies. Dates of the last year of growth from basic structural wood elements constrain house use, with the youngest tree-ring dated samples providing robust limiting ages, a terminus post quem (TPQ) on any construction activity or occupation.
Arctic archaeologists often maximize the database (see Mason and Rasic Reference Mason and Rasic2019) by relying on marine mammal or human bone shaped by the marine reservoir effect that ranges from several hundred to more than 1,000 years (Dumond and Griffin Reference Dumond and Griffin2002; Krus et al. Reference Krus, Jensen, Derek Hamilton and Sayle2019; Reuther et al. Reference Reuther, Shirar, Mason, Anderson, Coltrain, Freeburg, Bowers, Alix, Darwent and Norman2021). Local marine reservoir corrections (ΔR) are required for accurate calendrical calibration, and stable isotope measurements are needed to accurately estimate the contribution of marine carbon within sampled material (Coltrain et al. Reference Coltrain, Tackney and O’Rourke2016; Krus et al. Reference Krus, Jensen, Derek Hamilton and Sayle2019), including bone, hair, skin, or leather. Organic sediments or residues can be further offset by the incorporation of marine mammal oil (Morrison Reference Morrison1989; Park Reference Park1994; Smith et al. Reference Smith, Smith and Nilsen2018). This issue was recognized in the 1970s (Arundale Reference Arundale1981; McGhee and Tuck Reference Robert, Tuck and Moreau1976) in the eastern Arctic and remains a major consideration in the precise dating of archaeological events in coastal sites. These limitations have led Canadian archaeologists to preferentially assay “freshly” broken caribou bones (Friesen Reference Friesen2020; Morrison Reference Morrison1989; Nelson and McGhee Reference Nelson and McGhee2002).
In Alaska, a nihilistic dread took hold, as exemplified by the reluctance of the early 1980s Utqiaġvik Project to use even 14C dates, not to mention tree-ring dating (Hall and Fullerton Reference Hall and Fullerton1990), with a single exception (Sheehan Reference Sheehan1997:108). A shift in the dating paradigm followed the critique of Gerlach and Mason (Reference Craig and Mason1992), resulting in an enlarged database of radiocarbon dates that refine prehistoric cultural developments (Anderson et al. Reference Anderson, Brown, Junge and Duelks2019; Blumer Reference Blumer, Don and Richard2002; Jensen Reference Jensen2009; Mason and Bowers Reference Mason, Bowers and Grønnow2009; Mason and Rasic Reference Mason and Rasic2019). Today, dating efforts are benefiting from the use of Bayesian modeling (Brown et al. Reference Brown, Anderson, Junge and Duelks2021; Krus et al. Reference Krus, Jensen, Derek Hamilton and Sayle2019; Ledger et al. Reference Ledger, Forbes, Masson-Maclean, Charlotta Hillerdal, McManus-Fry, Jorge, Britton and Knecht2018) and improved radiocarbon laboratory procedures (Cook et al. Reference Cook, Ascough, Bonsall, Hamilton, Russell, Sayle, Scott and Bownes2015; Dunbar et al. Reference Dunbar, Cook, Naysmith, Tripney and Xu2016).
To accurately date the Birnirk and Thule occupations at the Rising Whale site, we obtained 50 radiocarbon and 15 tree-ring dates over four seasons of fieldwork from two multi-room dwellings. We used a Bayesian chronological framework to model these dates, investigating the overlap of occupations and the use life of each house. Subsequently, the set of calibrated ages was associated with the extensive, diagnostic assemblages of the twin polarities of the Birnirk-to-Thule transition (Morrison Reference Morrison1989, Reference Morrison2001) or the Punuk-to-Thule continuum (Harritt Reference Harritt2004; Yamaura Reference Yamaura and Allen1979).
Archaeological Context
The Rising Whale (KTZ-304) site on the low and short E-6 dune (Figure 2) represents the oldest occupation of the Inuit/Iñupiat tradition at Cape Espenberg. The small multicomponent site followed a hiatus of circa 300 years subsequent to a small and brief Ipiutak occupation on ridge E-8 between cal ADFootnote 1 600 and 750 that preceded the higher, wider, and longer (>200 m) ridge E-5 (Mason et al. Reference Mason, Hopkins and Plug1997). Dune E-6 formed by cal AD 700, establishing a potential lower limit of occupation (Mason et al. Reference Mason, Jensen, Rinck, Alix, Bowers and Hoffecker2020). At its initial settlement, this isolated dune—160 m in length and 25 m wide—was 2 m above a freshwater marsh and provided the only habitable space at the end of the spit, with a northward vista of low dunes and access to the open ocean. Houses were preferentially constructed on the dune crest (Figure 3), with other activity areas on a flat surface to the north, as indicated by GPR and magnetic survey (Urban et al. Reference Urban, Rasic, Alix, Anderson, Chisholm, Jacob, Manning, Mason, Tremayne and Vinson2019).
The dated samples are from two excavated semi-subterranean house features, separated by an unexcavated house depression (Figure 4). House F-12 consists, minimally, of three rooms connected by short passageways and through an elongated inner tunnel. It was constructed with well-preserved horizontally stacked logs, corner posts, planked floors, and alcoves. Lacking an internal hearth, it is associated with multiple external burning areas. Excavations revealed several stratigraphic and construction components with at least two superimposed wood floors (see Supplementary Text 1; Supplementary Figures 1–5). The artifact assemblage falls within late Birnirk, dominated by Natchuk and Thule II types open-socket harpoon heads and five other types (Figure 5).
Figure 4. Close-up map of KTZ-304 showing architectural elements of excavated features F-12 and F-21 (map by Laura Poupon, University Paris 1, 2017). (Color online)
Figure 5. Diverse Birnirk diagnostic harpoon head types uncovered in house F-12: (a) Okvik type, (b–c) Tuquok, (d) Natchuk, (e) Tipiruk, (f) Katoktok (g) Thule 4, (h) base fragment with three spurs, and (i) Tasik/Sicco (see Ford [Reference Ford1959] for harpoon head types’ photos; Claire Alix; line drawing, Sylvie Eliès, ArchAm-CNRS). (Color online)
House F-21 is a western Thule driftwood framed house, with a sub-rectangular main room, an attached side kitchen, and a 5 m entrance tunnel. Excavations identified several components in relation to construction sequence and stratigraphy, including one wood floor and two building phases in the passageway to the kitchen. The artifact assemblage is early Thule, with fewer harpoon heads, predominantly the Thule type II (Figure 6). The partially articulated bones of a female skeleton were found in two areas of the house (see Supplementary Text 1; Supplementary Figures 6 and 7). The excavation and analysis of these remains were approved following consultations led by the NPS with the locally affiliated tribal councils, which resulted in a signed Plan of Action that fulfilled the Native American Graves and Repatriation Act (NAGPRA) regulations.
Figure 6. Early Thule artifact types from house F-21: (a) open-socket harpoon head, (b) Thule type 2, (c) decorated winged needle case, and (d) closed socket self-bladed harpoon head (line drawings by Sylvie Eliès, ArchAm-CNRS).
Methods: Sampling and Chronological Modeling
We selected radiocarbon samples preferentially from contexts associated with domestic activities or from structural components in the two features, obtaining 50 AMS radiocarbon measurements on several sample types (Supplementary Table 1). Samples were submitted to five laboratories that maintain protocols of rigorous quality and assurance procedures and that participate in international inter-laboratory comparisons (Scott Reference Scott2003; Scott et al. Reference Scott2003, Reference Scott, Cook, Naysmith, Bryant and O’Donnell2007, Reference Scott, Gordon and Naysmith2010). We obtained three times more radiocarbon dates for F-12 (N = 37) than for F-21 (N = 13) both because of the greater complexity of F-12 architecture and a need to establish the age of its distinct archaeological levels and architectural and sedimentary units, especially in the north room and west excavation block.
We present the conventional radiocarbon ages (Stuiver and Polach Reference Minze and Polach1977) in Supplementary Table 1. The four closest points to Cape Espenberg in the Queen’s University Belfast’s Marine Reservoir Database (http://chrono.qub.ac.uk) were used to calculate a ΔR of 332 ± 66 to correct for local reservoir effects from marine carbon when necessary. We calculated calibrated date ranges with OxCal v4.4 using the relevant terrestrial and marine calibration curves of Reimer and others (Reference Reimer, Austin, Bard, Bayliss, Blackwell, Bronk Ramsey and Butzin2020) and Heaton and others (Reference Heaton, Köhler, Butzin, Bard, Reimer, Austin and Bronk Ramsey2020). Following the Cook and coauthors’ (Reference Cook, Ascough, Bonsall, Hamilton, Russell, Sayle, Scott and Bownes2015) correction of radiocarbon ages for material whose carbon derived from both terrestrial and marine sources, we used a mixed terrestrial and marine calibration curve that combines the internationally agreed calibration curves for terrestrial/atmospheric samples (IntCal20) with the calibration curve used for marine samples (Marine20). It is a modeled calibration because of the uncertainty associated with establishing the percentage of diet that derived from terrestrial/marine resources. Following Arneborg and coauthors (Reference Arneborg, Heinemeier, Lynnerup, Nielsen, Rud and Sveinbjörnsdóttir1999), we calculated the percentage of terrestrial protein consumed (with an uncertainty of 10%) using –12.0‰ and –20.0‰ as δ13C end members, where –20.0‰ equated to a 100% terrestrial diet and –12.0‰ represented a 100% marine diet.
In addition, we obtained 15 tree-ring dates from spruce (Picea sp.) structural members in F-12 (Supplementary Table 2), each of which was systematically mapped and collected (Supplementary Figures 3–5). Disk samples were prepared and several radii measured using a Velmex linear encoder and a laser measurement device with 0.001 mm resolution to produce raw ring width (RRW) series. All RRW were checked using the program COFECHA (Holmes Reference Holmes1983), and multiple radii measurements were averaged to produce a tree mean RRW for a given year. We cross-dated each mean RRW against each other using TSAP-Win software (Rinn Tech Reference Rinn2003) and built a highly sensitive floating chronology of 15 disk samples (Taïeb Reference Taïeb2023), which we cross-dated with the 1,000-year-long Kobuk River Master chronology available on the International Tree-Ring Data Bank (Zhao et al. Reference Alix, Mason, Norman and Lee2018).
We used a Bayesian approach in the interpretation of these chronological data following the procedures of Buck and others (Reference Buck, Cavanagh and Litton1996). Although simple calibrated dates provide accurate estimates of the age of samples, the ultimate aim of archaeologists involves establishing the dates of the behavioral events represented by the samples. In our case, we wanted to establish the timing of the various activities associated with each house: first its construction and then its subsequent long-term use(s). This objective goes beyond the age estimates from individual radiocarbon and dendrochronological samples. We estimated the temporal sequence of the activities within the houses through the Bayesian modeling of the absolute chronological data; we then incorporated the relative dating inferences provided by the house feature stratigraphy and inner feature relationships (Supplementary Figures 8 and 9). The probability estimates calculated by the chronological modeling are the posterior probabilities produced following Bayes’s rule and are not absolute (Bayes Reference Bayes1763; Kruschke Reference Kruschke2015). In turn, the chronological modeling results will likely change as further data become available, if modeled following alternate parameters, or both. The Bayesian modeling was applied with the program OxCal v4.4, which uses Markov chain Monte Carlo sampling techniques. Amodel agreement index calculated by OxCal was used to evaluate the fit between the OxCal model and data, with values ≥60 indicating good agreement (Bronk Ramsey Reference Bronk Ramsey2009).
The structures of the Bayesian chronological model are fully described in Supplementary Text 1, and the code is provided in Supplementary Text 2. The archaeological context of each radiocarbon sample and the stratigraphic relationships between samples are described in Supplementary Text 1. To create our model in OxCal, we placed the radiocarbon dates from F-12 and F-21 into two independent phases (represented by the Phase() function) for each structure and nested phases corresponding to the archaeological contexts and stratigraphy (Supplementary Text 1). We used the Boundary() function in OxCal to estimate the start and end date of activity associated with each structure (see Bronk Ramsey Reference Bronk Ramsey1995, Reference Bronk Ramsey1998, Reference Bronk Ramsey2001). All absolute dates obtained from driftwood samples were modeled as TPQ to account for the time lag between the cellular death of these samples and their deposition into archaeological contexts.
When multiple radiocarbon measurements dated the presumed identical archaeological context, we used chi-square tests to assess their consistency (Ward and Wilson Reference Ward and Wilson1978). In cases where multiple radiocarbon measurements from the same archaeological context derived from both marine and terrestrial sources, we used the Combine() function in OxCal to provide Acomb test statistics to further assess consistency (Bronk Ramsey Reference Bronk Ramsey1995, Reference Bronk Ramsey2009, Reference Bronk Ramsey2017). Sample descriptions and chi-square test results can be found in the supplemental material (Supplementary Text 1; Supplementary Tables 3 and 4). Dated samples from fill contexts that fail these chi-square tests or ones not paired with another dated sample from a fill context were also modeled as TPQ because it is feasible that the matrix and samples in these contexts were redeposited.
Results
Radiocarbon and Tree-Ring Ages
As could be expected, radiocarbon assays on wood are systematically older than radiocarbon dates on caribou bones, except for the unidentified wood fragment (Beta-220021) obtained prior to house excavation (Supplementary Table 1).
Radiocarbon dates on well-contextualized caribou bones in each house cover a range of about 100 radiocarbon years, except for one early date on a sample in F-21 that came from an upper fill level (Supplementay Table 1). In general, the radiocarbon ages are consistent within each house features but do not strictly reflect sample stratigraphic placement. The age of a human mandible recovered in the sod layer outside Birnirk house F-12 provides an upper limiting age, cal AD 1220–1670 (p = 0.95), which is certainly younger than the house’s main occupation (Supplementary Text 1; Supplementary Figure 8), despite its large uncertainty. This contrasts with the age of a rib from the skeleton interred within Thule house F-21, which falls within the timing of the occupation (Supplementary Text 1).
Dates on well-contextualized structural wood in F-12 extend over 240 radiocarbon years, being as old as cal AD 770–980 (OS-96128, p = 0.95) and as young as cal AD 1030–1170 (OS-96069, p = 0.95). In contrast, tree-ring dates (Supplementary Table 2) span just over 100 years, with the oldest end-ring date at AD 1080 (12w126-05) and the youngest at AD 1186 (12w113-09). Eight of the 15 structural timber tree-ring dates were used in building the walls and lowest floor of F-12 north room and represent the original construction of the house. These dates also span over 100 years: the oldest is the lowest horizontal log of the north wall (12w126-05, AD 1080), and the youngest is one horizontal log of the lowest floor (12w113-09, AD 1186). A tree-ring date at AD 1177 is from a post in the East room that is only 20 years from the youngest date. This latter date, together with the overlay of all structural elements into one footprint, suggests one main building episode for this house frame (Supplementary Figure 5; Supplementary Table 2). The youngest of all the tree-ring dates provides a TPQ for the earliest possible construction of the house,with all cultural activities occurring afterward.
Bayesian Modeling
The Amodel agreement value for the chronological model is greater than 60, suggesting good overall agreement between the model assumptions and the absolute chronological data. Supplementary Tables 3 and 4 present the full details of the posterior probabilities estimated from chronological modeling.
The construction and domestic activities within Birnirk house F-12 likely began after the youngest end ring date (12w113-09, AD 1186); the primary Bayesian model suggests that this start date may have been as late as cal AD 1205 (95% probability; Figure 7; Primary Model: start of Feature 12). This model suggests that activity associated with F-12 ended between cal AD 1210 and 1260 (95% probability; Figure 7; Primary Model: end of Feature 12) and spanned between 15 and 110 years (95% probability; Figure 8, Primary Model: Feature 12 span). Subsequently, the model estimates that the construction and occupation related to Thule house F-21 began cal AD 1265–1300 (95% probability; Figure 7; Primary Model: start of Feature 21), 20–80 years following the termination of the occupation within F-12 (95% probability; Figures 8, Primary Model: Gap between Feature 12 and 21). Activity related to F-21 is estimated to have ended in cal AD 1285–1325 (95% probability; Figure 7; Primary Model: end of Feature 21) and possibly spanned only 1–50 years (Figure 8, Primary Model: Feature 21 span). Additionally, the model suggests that F-12 was likely occupied for a longer period than F-21 (Figures 8; Supplementary Table 4), and there is no overlap in the occupation of the two houses.
Figure 7. Posterior probabilities for the estimated start and end dates of house occupations from the Bayesian models (dotted line indicates the TPQ for F-12 from tree-ring date on sample 12w113-09).
Figure 8. Posterior probabilities for the estimated timespans from the Bayesian models.
Age estimates from the outlying extramural activity areas of Birnirk house F-12 fit reasonably with the age of the structure, indicating contemporaneity between extramural activity areas and that of the house (Supplementary Text 1). Likewise, the date of the burned area connected to Thule house F-21 falls in the interval modeled for the remainder of that house (Supplementary Text 1).
An anonymous reviewer for an earlier version of this article suggested that we create an alternate version of this model for comparison to assess the influence of the absolute dates from wood samples on the results. We conducted a sensitivity analysis for comparison by removing all absolute dates from wood samples, so this alternate model would consider only radiocarbon measurements from bone samples. Other than removing the absolute dates from wood samples, the alternate model is virtually identical to the primary model. The alternate model also produced good Amodel agreement, and the results were nearly identical to those derived by the primary model (Supplementary Tables 3 and 4).
Discussion
The implications of our analyses are threefold. First, we precisely date two adjacent houses, establishing a lack of contemporaneity; second, we demonstrate the power of combining tree-ring and radiocarbon dating; and third, we place the Birnirk-to-Thule transition in the second half of the thirteenth century.
House Contemporaneity
Our series of dates indicate a caesura of occupation of between 20 and 80 years (95% probability; Figure 8, Primary Model: gap between Feature 12 and 21) from the end of occupation in the late Birnirk house to the construction and use of the nearby early Thule house. One circumstance is striking: no evidence exists that the succeeding Thule inhabitants extracted structural wood from the older house. At a minimum, this suggests that the Birnirk house was either purposefully avoided for use as a living structure and as construction material, or it was sufficiently filled by sand to render it invisible or unusable, as indicated in a cover sand bed. In any case, the Birnirk residents of the Rising Whale site were likely absent from the area, allowing the Thule group to settle. The alternative ramifications of a decadal- versus a century-long hiatus are tremendously significant in view of the contrasting architectural and assemblage character of the two houses, which contain unmixed artifact assemblages that encapsulate the transformative process leading to the preeminence of the Inuit tradition (see Mason Reference Mason2020:12–16). This situation is unlike nearly all northwest Alaska locations that typically witnessed builder disturbance (i.e., digging) into older deposits during house construction.
The surprising absence of contemporaneity for houses built so close to one another strengthens the idea that site occupation, even on Alaskan beach ridges, was more complex than commonly believed. Houses should be evaluated individually to establish the chronologies of successive occupations within them (Ledger et al. Reference Ledger, Forbes, Masson-Maclean, Charlotta Hillerdal, McManus-Fry, Jorge, Britton and Knecht2018) or within sites (see Shirar Reference Shirar2011:9, 13). Increased precision in dating house occupation(s) will refine the temporal pattern of landscape use, establishing sequences of occupations and abandonment, a standard practice in other regions (see Cameron and Tomka Reference Cameron and Tomka1993; Jazwa et al. Reference Jazwa, Gamble and Kennett2013).
Determining the simultaneity of house occupations at Thule sites across the North American Arctic remains a key issue in the contentious debate over the timing and demography of Thule migrant populations (Friesen Reference Friesen2020; Mason Reference Mason2020; McGhee Reference McGhee, Maschner, Mason and McGhee2009; Morrison Reference Morrison1999; Park Reference Park1997). In many cases, contemporaneity is assumed solely on the basis of house proximity, artifact assemblage uniformity, or both (e.g., Ford Reference Ford1959; Giddings Reference Giddings1952; Giddings and Anderson Reference Giddings and Anderson1986; Park Reference Park1997; Whitridge Reference Whitridge, Frink, Shepard and Reinhardt2002). In Kotzebue Sound, even when site complexity is acknowledged, archaeologists regularly use the surface contours of houses on discrete beach ridges to infer contemporaneous occupations, relying on a handful of radiocarbon ages (see Darwent et al. Reference Darwent, Mason, Hoffecker and Darwent2013:443). The resulting chronology leads to a simplified and misleading evolutionary and demographic sequence (see Schaaf Reference Schaaf1995; Shirar Reference Shirar2011). This approach proved a chimera because so little is known of Thule seasonal settlement pattern and landscape use, with even less known for the Birnirk occupations that lack adequate contextual data.
The Importance of Coupling Tree-Ring and Radiocarbon Dates
Our precise chronology for the Birnirk occupation was enabled by tree-ring dates from house structural elements and radiocarbon ages of broken caribou bones and end rings of wood samples. The tree-ring results were crucial in this effort in which the calibrated radiocarbon dates fell on a plateau within multiple acceptable ranges of the calibration curve between cal AD 1050 and AD 1220. By contrast, the radiocarbon assays from F-21 fall on a steeper section of the calibration curve, yielding a tighter calendar age.
In the Arctic, many archaeologists continue to assay charcoal fragments for dating sites less than 2,000 years old, often for want of more reliable types of material (e.g., Anderson and Freeburg Reference Anderson and Freeburg2013; Freeburg and Anderson Reference Freeburg and Anderson2012). Dating charcoal fragments, which most likely originate as driftwood, implicitly incorporates old wood uncertainties in Arctic sites. As an antidote, the assay of wood and driftwood was rejected by a generation of Canadian archaeologists (Ledger et al. Reference Ledger, Forbes, Masson-Maclean, Charlotta Hillerdal, McManus-Fry, Jorge, Britton and Knecht2018; Morrison Reference Morrison1989) because of the uncertain lag time between tree death and the use of driftwood for fuel or technology. A reactionary bias against dating any wood can be misguided and simplistic, because the uncertainty in a calibrated radiocarbon assay is usually on the order of a century or more. By contrast, tree-ring-dateable wood can offer greater precision, even accounting for the time lag between wood delivery, its selection, and use (Alix Reference Alix, Friesen and Mason2016). In addition, as shown in this study and in Griggs and others (Reference Carol, Kocik, Urban, Manning, Douglas and Wanni2019) on the Kobuk River, the combination of radiocarbon and tree-rings of well-contextualized structural timbers yields a robust methodology to understand house construction and renovation and to constrain the range of house activities. Precise dating and the details of building sequences are more firmly established by the critical selection of dating material (see Bayliss Reference Bayliss2009) and the systematic recording of house architecture (Alix et al. Reference Alix, Mason, Norman and Lee2018; Friesen and Méreuze Reference Friesen and Méreuze2020; Norman et al. Reference Norman, Max Friesen, Alix, O’Rourke and Mason2017).
The radiocarbon measurements from caribou bones are thought to provide good chronological proxies for the use of the house, whereas the tree-ring dates and radiocarbon measurements from wood provide important chronological data on house construction and modification. In the end, the youngest tree-ring dates from Feature 12 fall right within the probability interval for the start of activity associated with this structure, in both the primary and the alternate Bayesian models, highlighting the strength of the approach. Hence, one major implication of our results is a call for a systematic consideration of well-preserved structural wood remains in Arctic sites (Alix Reference Alix, Friesen and Mason2016; Taïeb et al. Reference Taïeb, Daux, Alix and Hatté2023). Driftwood may complicate the assembly of tree-ring sequences, but dendrochronology, in addition to its potential for climatic studies, provides the crucial contextual data to follow the “life history” of buildings and, ultimately, of the people who lived in them.
Implications for the First Presence of Thule in Alaska
Sixty years ago Ford (Reference Ford1959) hypothesized that the Birnirk-to-Thule transition occurred in the mid-thirteenth century. Based primarily on trait comparisons, he proposed that “the Ahteut site [on the Kobuk River], with a tree-ring date of 1250 AD [was] coeval with the end of the occupation of the Birnirk site and the beginning of occupation of the Nunagiak site in the Point Barrow area and also with the Ievoghiyoq [Ayveghyaget] site on St Lawrence Island” (Ford Reference Ford1959:243). Taylor (Reference Taylor1963) offered a similar insight from the Canadian Arctic, whereas 40 years later, Blumer (Reference Blumer, Don and Richard2002:75) collated St. Lawrence Island data to conclude that Thule succeeded Punuk between cal AD 880 and 1300; this interval refined by a 14C date on a kayak element from Ayveghyaget House 7 to within the thirteenth century (Anichtchenko Reference Anichtchenko2016). A thirteenth-century occupation was confirmed at the nearby Mayughaaq mound where Bandi and Bürgi (Reference Bandi and Bürgi1975) had dated wood from a Punuk structure to within cal AD 1220–1360. In Norton Sound at Cape Denbigh, Murray and others (Reference Murray, Robertson and Ferrara2003) assayed diagnostic harpoon heads from the Nukleet midden, also placing the earliest age of the Thule component in the thirteenth century.
Our dating of the two houses at Cape Espenberg indicates that the Birnirk-to-Thule transition occurred as much as two centuries before the widely accepted circa AD 1000 datum for northern Alaska (Mason Reference Mason, Friesen and Mason2016; Morrison Reference Morrison2001): this implies Thule, as a homogenized cultural manifestation derived from Birnirk and Punuk, was absent from coastal Alaska before about AD 1250. This dating confirms Ford’s (Reference Ford1959) prognostication, as well as data from across Bering Strait, especially Blumer (Reference Blumer, Don and Richard2002:93) for St. Lawrence Island and Bronshtein and others (Reference Bronshtein, Dneprovsky, Savinetsky, Friesen and Mason2016) for northern Chukotka, that place the transition between cal AD 1200 and 1400 (see also Dneprovsky Reference Dneprovsky2006).
On the Alaskan coast, evidence for the Birnirk-to-Thule transition is either less precise, absent, or equivocal (Mason Reference Mason, Friesen and Mason2016:497). Close to Cape Espenberg, two adjacent excavated houses at Deering, showing strong artifact similarities with F-21 and in some cases F-12, fall mostly within the thirteenth century (Mason and Bowers Reference Mason, Bowers and Grønnow2009). Across the sound from Cape Espenberg, the Birnirk–Thule sequence is identified on the Cape Krusenstern beach ridges, but the few legacy ages are too imprecise (Giddings and Anderson Reference Giddings and Anderson1986) and recent extramural soil probes yield little clarity (Anderson and Freeburg Reference Anderson and Freeburg2013; Freeburg and Anderson Reference Freeburg and Anderson2012). Farther north, Walakpa with its stratigraphic Birnirk–Thule succession is also dated with only two “acceptable” overlapping carbon solid ages on charcoals fragments (Morrison Reference Morrison2001), and Piġniq, the Birnirk type site, does not document the transition despite some new dates (Anichtchenko Reference Anichtchenko2016; Brown et al. Reference Brown, Anderson, Junge and Duelks2021; Clark et al. Reference Clark, Horstmann, de Vernal, Jensen and Misarti2019; Petherick et al. Reference Petherick, Reuther, Shirar, Anderson and DeSantis2021; Unkel et al. Reference Unkel, Norman, Tackney, Krus, Jensen, Alix, Mason and O’Rourke2022).
Given this lack of contextualization of radiocarbon dates across the region, the ages of the two houses at the Rising Whale site are important for establishing the first presence of Thule on the Alaskan coast. The Birnirk or Punuk/Birnirk to Thule sequence can be authenticated there, within a younger and constrained time set, in the second half of the thirteenth century. This is at the same time as Giddings’s (Reference Giddings1952) proposed time period—post-AD 1250—for the first entry of Thule into the interior northwest Alaska, and it is very much aligned with the now-accepted thirteenth-century migration into western Canada (Friesen and Arnold Reference Friesen and Charles2008). Early Thule Western Canadian Arctic sites, such as Nelson River, Cache Point, and Tiktalik, have dates that fall into the thirteenth to fourteenth century, with the same issues of long calibrated intervals as described throughout our article (Friesen and Arnold Reference Friesen and Charles2008; Morrison Reference Morrison, Maschner, Mason and McGhee2009). Our results are in agreement with Friesen and Arnold’s (Reference Friesen and Charles2008:534) calculation of a brief early Thule period and call into question the length of the Thule residence in Alaska before they traveled eastward. Did the early Thule inhabitants of house F-21 at Cape Espenberg explore the northwest interior while others moved along the north coast of Alaska? Or were people established in the north (at Piġniq and vicinity) before people settled at Rising Whale KTZ-304? A Birnirk-to-Thule transition in Alaska within the thirteenth century with nearly contemporaneous migrations in the interior and farther east to the Canadian Arctic and Greenland supports a population on the move from the Siberian coast through Alaska (Mason Reference Mason2020).
We recommend continuing to carefully sample and date or redate well-contextualized material from coastal sites such as Piġniq and Kugusuguruk (Figure 1) and other sites from interior and northern Alaska to test whether their occupations precede that of F-12 and F-21 and so replicate the rejuvenation of the Birnirk-to-Thule transition and first Thule presence in continental Alaska. The social complexity of the broader Bering Strait coasts involved numerous hybrid communities, as envisioned by Ackerman (Reference Ackerman and John1962) and Arutiunov and Bronshtein (Reference Arutiunov and Bronstein1985). Accepting that complexity, a detailed and complete description and analysis of artifact types and technology for both houses is ongoing that will be essential in associating this precise dating with peoples’ movements, activities, exchange networks, and cultural preferences.
Conclusion
The Rising Whale site, with its successive late Birnirk and early Thule occupations, brackets the transition between the two archaeological phases. This tightly dated but not overlapping succession of two houses can be dated to after the mid-thirteenth century, a late date for the transition but one that is nevertheless in accord with the Bering Strait regional record and earliest well-contextualized dates of the Canadian Arctic and Greenland Thule sites (Gulløv and McGhee Reference Gulløv and McGhee2006; McGhee Reference McGhee, Appelt, Berglund and Gulløv2000, Reference McGhee, Maschner, Mason and McGhee2009; Morrison Reference Morrison, Maschner, Mason and McGhee2009).
Our results are unique in their precision at the house level, thereby enriching our understanding of this key moment in the cultural development of northwestern Alaska. Our findings indicate the need for renewed analyses of specific contexts that take taphonomic processes into account (Mayeux et al. Reference Mayeux, Alix, Mason, Bigelow and Petit2024), thereby enabling a clearer understanding of occupation succession. Finally, our analysis and results demonstrate the power of Bayesian analysis and of combining tree-ring dating of well-contextualized architectural timbers with radiocarbon dates on terrestrial mammals and other organic materials.
Acknowledgments
Our deepest thanks go to the community of Shishmaref, especially the families who have ancestral ties to Cape Espenberg, for supporting our archaeology efforts and agreeing to participate in the project. Fred Goodhope Jr. with his son and daughter assisted with the excavation of the burial. We extend our acknowledgements to the community of Deering and Kotzebue and to everyone who had a role in the project. We thank John F. Hoffecker (INSTAAR, University of Colorado Boulder), Christyan M. Darwent, and John D. Darwent (University of California Davis) for their efforts in 2010–2011, as well as all members of the crews who during each summer fieldwork enabled the careful excavation of the two houses at KTZ-304 Rising Whale site. At the University of Alaska Fairbanks (UAF), we thank Glenn P. Juday, Ryan Jess, and Mike Lorain for access to the UAF Tree-Ring Lab and help with measuring disk samples, as well as Nancy H. Bigelow at the Alaska Quaternary Center (AQC) and Chris Maio at the Arctic Coastal Geoscience Lab. We also thank Sylvie Eliès of CNRS-UMR8096 for artifact illustrations and house feature maps, Jeff Rasic at the National Park Service (NPS), Amber Lincoln (LA Natural History Museum), Susanne Grieve-Rawson at Heritage Preservation Field Support Solutions, and Anna Kerttula for her dedication to our project and Arctic research in general. All research and project activities were conducted under a Memorandum of Agreement between the NPS, the National Science Foundation (NSF), and the Alaska State Preservation Officer and a NAGPRA Plan of Action between the NPS, the Native Village of Shishmaref, and the Native Village of Deering, which included approval to excavate human remains and conduct radio-isotopic analyses on the excavated remains (ARPA Permits 16-BELA/AKRO-001, 17-BELA/AKRO-02, 18-BELA/AKRO-01 and BELA-2016-SCI-001 and BELA-2017-SCI-003).
Funding Statement
This research was supported by the National Science Foundation, Office of Polar Programs (OPP) Arctic Social Sciences under Grants ARC-0755725 and collaborative Grant ARC-1523160, ARC-1523205, ARC-1523059, and ARC-1523059. Field research was also supported by the Archaeology Commission of the French Ministry of Foreign Affairs, the Chaire Paris 1 Pantheon-Sorbonne University/CNRS awarded to Claire Alix, CNRS research unit Archéologie des Amériques (UMR 8096), and the AQC in the College of Natural Science & Mathematics at UAF.
Data Availability Statement
Data and collections used in this research were on loan from the National Park Service at the University of Alaska Fairbanks and are now available at the NPS Alaska Regional office in Anchorage.
Competing Interests
The authors declare none.
Supplementary Material
The supplementary material for this article can be found at https://doi.org/10.1017/aaq.2025.21.
Supplementary Text 1. Feature and sample description with Chi-Square results.
Supplementary Text 2. Code for OxCal Models.
Supplementary Figure 1. Map of F-12: Under sod level (map: Sylvie Eliès and Claire Alix).
Supplementary Figure 2. Map of F-12: Roof level and external areas (map: Sylvie Eliès and Claire Alix).
Supplementary Figure 3. Map of F-12: Below Roof level (map: Sylvie Eliès and Claire Alix).
Supplementary Figure 4. Map of F-12: First planked floor level (map: Sylvie Eliès and Claire Alix).
Supplementary Figure 5. Map of F-12: Second (lowest) floor level (map: Sylvie Eliès and Claire Alix).
Supplementary Figure 6. Map of F-21: Floor level (map: Sylvie Eliès, Claire Alix, and Lauren Norman).
Supplementary Figure 7. Map of F-21: Below Floor level (map: Sylvie Eliès, Claire Alix, and Lauren Norman).
Supplementary Figure 8. Matrix of F-12 radiocarbon and tree-ring dates.
Supplementary Figure 9. Matrix of F-21 radiocarbon dates.
Supplementary Table 1. List of radiocarbon dates for Features F-12 and F-21.
Supplementary Table 2. Tree-ring dates from house F-12.
Supplementary Table 3. Posterior probabilities from the Bayesian models for the estimated start and end dates for the house occupations.
Supplementary Table 4. Posterior probabilities from the Bayesian models for estimated timespans.
Open access
