Dating and Digging Stratified Archaeology in Circumpolar North America : A View from Nunalleq , Southwestern Alaska

Through the case study of the Thule-era village site of Nunalleq (GDN-248), this paper presents 14C dating results and perspectives on the issues associated with radiocarbon dating stratified archaeological sites in circumpolar North America. The objective was to investigate relative variation in the 14C age of ecofacts with the aim of establishing a hierarchy of dating suitability for Nunalleq that could more widely inform 14C sample selection on archaeological sites across the North American sub-Arctic and Arctic and Greenland. Owing to the complexities associated with interpreting and establishing the relative chronology of the deeply stratified sod deposits at Nunalleq, we adopted open-area excavation and single-context recording methods. This approach, we suggest, allowed us to eliminate stratigraphic complexity as a source of variation in 14C measurements and to assess the taphonomic issues associated with dating different ecofacts. In total, 16 samples were submitted for dating, comprising two sets of eight different ecofacts, one from each of two stratigraphically contemporary but spatially discrete contexts. In most instances, the 14C ages of ecofacts were statistically indistinguishable between the two contexts and support the relative chronological relationships established by excavation. Only Elymus arenarius (grass) manufactures and Heleomyzidae (fly) puparia produced different ages in the two contexts, variations that suggest that these items are unreliable dating materials. As noted in previous studies, Phoca sp. (seal) and Oncorhynchus sp. (salmon) bone collagen demonstrated a strong marine reservoir effect (c. 700 14C yr.). Picea sp. (wood chips) were marginally older than seeds from edible berries (Rubus chamaemorus and Empetrum nigrum) and caribou (Rangifer tarandus) bone collagen, which provided the most consistent ages.


INTRODUCTION
Accurate and robust radiocarbon chronologies from Thuleera and antecedent archaeological sites in the North American Arctic and sub-Arctic (from Alaska in the west to Greenland in the east) are rare (McGhee, 2000).Despite the excellent preservation potential for organic materials in the permafrost soils of circumpolar regions, researchers have often struggled with constructing what they consider to be accurate radiocarbon-based chronologies (e.g., Gerlach and Mason, 1992;Park, 1994;Grønnow and Jensen, 2003;Betts, 2008;McGhee, 2009;Anstey et al., 2016), and the site of Nunalleq, the subject of this paper, is no exception.The difficulties are much discussed, and numerous authors have contributed to the debate (e.g., Arundale, 1981;Morrison, 1989;Gerlach and Mason, 1992;Park, 1994;Dumond and Griffin, 2002;West, 2011;Anderson and Freeburg, 2013;Ramsden and Rankin, 2013).A primary cause of dating issues-although unrelated to radiocarbon dating itself-stems from the cold and dry nature of northern environments, which results in limited biological activity and in many places almost non-existent soil formation (McGhee, 1996).This general dearth of soil accumulation results in a widespread, although not complete, absence of stratified archaeology in the North.Stratigraphy, and the relative chronological information contained within, is fundamental to dating in archaeology.It provides the means to differentiate occupation episodes, estimate their duration, and perhaps most importantly, assess the reliability of other chronological information such as 14 C data.Where there is little soil accumulation, traces of multiple and chronologically distinct cultural episodes may form only a thin archaeological deposit, giving the impression of a single short-term occupation.However, 14 C dates from such settings often reveal palimpsests of multiple cultural events, and there is no way of establishing the association between different artefacts or features and their absolute age (e.g., Grønnow and Jensen, 2003).The widespread absence of stratigraphy in the North has therefore deprived archaeologists of a powerful route to building robust chronologies for the history of Arctic peoples (c.f.Giddings, 1967).
The major problems associated with radiocarbon dating can be broadly summarized as taphonomic factors.Preeminent in this list is the marine reservoir effect (McGhee and Tuck, 1976;Park, 1994).Marine resources were central to traditional lifeways in circumpolar North America, and products of the sea are abundant on archaeological sites (Dumond, 1987;Woollett et al., 2000;Grønnow and Jensen, 2003).Consequently, a large proportion of the organic remains available for radiocarbon dating are either marine in nature or have come into contact with marine resources (Morrison, 1989;Park, 1994;Gulløv, 1997).In fact, the abundance of marine products and their associated oils has anecdotally been implicated as an auxiliary factor contributing to the difficulties experienced in radiocarbon dating Arctic sites.It has been suggested that seal oils permeate everything to such an extent that the radiocarbon ages of terrestrial materials are significantly skewed (Park, 1994).Although this idea is widespread in the literature (e.g., Morrison, 1989;Park, 1994;McGhee, 2000;Anderson and Freeburg, 2013), the marine mammal oil effect has yet to be demonstrated empirically.
A further, somewhat diffuse, phenomenon is something that Park (1994:31) termed the "terrestrial reservoir effect."This label has been used to group diverse situations in which 14 C assays on materials that are exclusively derived from the terrestrial biosphere-i.e., terrestrial plants or herbivore bone collagen-produce what are deemed to be aberrant ages (Morrison, 1989;Park, 1994).For example, terrestrial (typically woody) plant macrofossils that have been interpreted as being contemporary sometimes provide very different 14 C ages (e.g., McGhee and Tuck, 1976;Schledermann and McCullough, 1980;Jensen, J.F., 2006).Similar problems have also been noted in Arctic lake sediments, where differing parts of terrestrial plants (e.g., wood, leaves, and seeds) recovered from the same depth in a lake can have sharply differing 14 C activities (Mann et al., 2002;Oswald et al., 2005).Varying reasons for this anomaly have been advanced.These include 1) slow organic decay in the Arctic that results in longer residence times in permafrost soils for "old" carbon, which then becomes incorporated into plant and herbivore tissues; 2) differential 14 C fractionation between plant species (Morrison, 1989;Park, 1994), although modern radiocarbon laboratories correct for this problem (Walker, 2005); and 3) taphonomic issues that result in certain macrofossils having an "inbuilt" age (sensu McFadgen, 1982;i.e., the time differential between the age of death and the date of deposition).
A less explicitly discussed problem (although see Morrison, 1989) may arise as a result of the challenging logistics associated with fieldwork in Arctic regions.Sites are often difficult and expensive to access, resulting in small-scale investigations by a limited number of excavators or non-specialists working under time pressure.Consequently, when excavating, Arctic researchers have tended to adopt the planum or Wheeler box-grid methods or hybrids of the two (e.g., Knecht and Jordan, 1985;Sutherland, 2009; although see Savelle and Habu, 2004;Hodgetts et al., 2015 for exceptions).Indeed, the initial rescue excavations at Nunalleq were conducted in this manner.The defining feature of these methods is digging in a series of independent grid squares.In each square, the excavator records changes (with reference to stratigraphic sections) at either arbitrary or natural levels, thereby establishing a relative chronological sequence for that square.The associations of deposits excavated in different squares are then inferred between these islands of information.Both approaches afford lowcost, rapid excavation of large areas (Renfrew and Bahn, 2004); however, they can result in imprecise appreciation of deep and complex site stratigraphy and a tendency to summarize it (Branch et al., 2005).Relative chronologies can be established with certainty for both deposits and artefacts and ecofacts within a single square, but temporal relationships between squares can potentially be misinterpreted.
If radiocarbon dates are then procured on organic materials drawn from multiple squares, this may pose interpretational problems.Indeed, the initial radiocarbon rangefinder data from Nunalleq, which contained age/depth inconsistencies (discussed below), stands as an example.Strong understanding of the sequence of events that led to the formation of archaeological deposits is essential for producing both relative and radiometric-based chronologies of a site (Bayliss, 2009).Without this understanding, the taphonomic issues identified above, which are known to result in erroneous 14 C ages, can become conflated with imprecision in stratigraphic interpretation to the point that these two sources of error are seemingly inseparable.Therefore, in order to understand the cause of variations in the 14 C age of different dating materials, it is essential to be certain that relative chronological differences are not the source of age variation.
A valuable method for evaluating the taphonomic issues associated with dating different ecofacts and artefacts is to undertake multiple radiocarbon measurements on materials of contemporary age.Unfortunately, because of a general or presumed paucity of ecofactual remains (e.g., Jensen, A.M., 2006) on Arctic archaeological sites, few investigations of this nature have been undertaken.Where such work does exist, it has tended to focus on elucidating the strength of the marine reservoir effect (e.g., Dumond and Griffin, 2002).The purpose of this paper is to use the Thule-era site of Nunalleq, which contains a plethora of different ecofactual remains, to expand the scope of such studies.Using open-area excavation and single-context recording, we identified two spatially discrete yet contemporary house floors that could be used to investigate the 14 C age variation of a series of common and rarely examined ecofactual remains.Using these data, we aim to (1) test the accuracy of single-context recording as a means of establishing relative temporal relationships of deposits on Arctic archaeological sites; 2) use 14 C data to investigate the taphonomy of ecofactual remains present within sod dwellings and determine which ecofact(s) provide the most reliable 14 C ages in such situations; and 3) establish a hierarchy of suitability for the dating of ecofacts from Nunalleq that can inform dating decisions on other archaeological sites in the New World Arctic.

Nunalleq
Nunalleq (GDN-248) is located close to the Yup'ik Eskimo village of Quinhagak (59˚45′12″ N, 161˚54′10″ W) on the coast of the Bering Sea in southwestern Alaska (Fig. 1).The site is set within the organic-rich delta of the Kanektok and Arolik Rivers, which form part of the wider braided system of the Yukon-Kuskokwim (Y-K) Delta, a landscape characterized by flat, low-lying treeless tundra dotted with lakes and underlain by discontinuous permafrost (Alaska Center for Conservation Science, 2016).The archaeological site was discovered in 2009 by residents of Quinhagak, who found a series of artefacts on the beach near the site.Rescue excavations of the western part of the site using the planum method were initiated that year in conjunction with the local community and continued in 2010.The result was the discovery of the remains of a multiphase village with numerous house floor layers that yielded tens of thousands of in situ lithics, pottery fragments, and artefacts of wood and grass.In addition, there were exceptionally well preserved ecofacts, including bone from terrestrial and marine animals, diverse plant remains, insects, fur, and human hair (Britton et al., 2013;Forbes et al., 2015).A suite of rangefinder dates on materials excavated between 2009 and 2012 returned ages from 182 ± 37 BP to 650 ± 40 BP (Table 1), implying a Thule-era occupation of three to four centuries (Britton et al., 2013).A simple Bayesian model combining these data suggested occupation dating from cal.AD 1310 to AD 1420, with abandonment somewhere between AD 1650 and AD 1750 (Britton et al., in press), which is consistent with the absence of Euromerican material culture at the site.Although the organic remains to be dated were selected exclusively from short-lived terrestrial species, the resulting data presented some inconsistencies.Contrary to expectations, the youngest 14 C age was returned on material excavated from one of the deepest levels of the site, and materials from the same level but different squares produced significantly varying ages (Fig. 2).In one instance, three assays from the same level and square (Square 1, Level 4) returned a series of statistically different dates (Table 1).A further issue was apparent in Square 2, where the 14 C assay from level 3 predated the assay from level 4. Importantly, Squares 1 and 2 furnished the oldest radiocarbon dates from the site and suggest that the beginning of occupation is unlikely to have pre-dated AD 1450.
Owing to a growing appreciation of the complexity of the stratigraphy (borne out in the inconsistencies of 14  oldest of which is yet to be excavated in its entirety.Each is differentiated by episodes of architectural remodeling of the larger sod structure.The occupational deposits are sealed by charred sod and collapsed structural remains that may relate to conflict connected to the Bow and Arrow Wars (Funk, 2010;Fienup-Riordan and Rearden, 2016).Resting above these layers are further sod deposits linked with the long-term collapse of the structure (Figs. 3 and 4).

Excavation and Sampling Methods
Samples analyzed in this study were collected from archaeological deposits excavated and recorded using the single-context recording system.This system follows the principle that each stratigraphic unit (defined as a context) is identified, exposed in its entirety, and recorded before being removed (Branch et al., 2005).In practice, this meant that deposits at Nunalleq were excavated in reverse chronological order, with the uppermost (most recent) stratigraphic layer being fully exposed across the excavation grid, assigned a unique context number, recorded (in plan form with measurements of levels), and then sampled and removed to reveal the next deposit.A Harris matrix was constructed to represent the stratigraphic relationships between each excavated context and the next, allowing the establishment of the relative chronologies of all archaeological deposits at the site (Harris, 1979).
Artefact and large animal bone provenance was identified by recording the context number and the grid square in which they were found.In addition, two sets of sampleslarge (2 L) general biological analysis (GBA) and bulk (18 L) faunal samples (Dobney et al., 1992)-were collected from each context, using clean tools, and placed into heavy-gauge plastic bags.In the case of floor layers, GBA samples were taken from each grid square (i.e., multiple samples from single contexts).These were then transported to the University of Aberdeen and stored at 4˚C.Bulk faunal samples were water screened in Quinhagak using a 3 mm mesh.

Context Selection for This Study
Two contemporary, although spatially discrete, house floors from the terminal phase of occupation at Nunalleq (contexts 14035 and 14065; Figs.3; 4A) were selected for study.On the basis of the relationships established through excavation, namely that the floors were buried directly below structural debris and charred sod or wood (Fig. 4B,  C), these deposits were interpreted as representing contemporary occupation surfaces.Both contexts were relatively thin, ranging in thickness from 4 to 8 cm (14035) and 5 to 9 cm (14065).These deposits represent the terminal levels of a sequence of at least nine distinct house floors of more substantial accumulation, which range in thickness from 8 to 12 cm.A sondage (boring or drilling) in the western portion of the site suggests that a further 40 cm of deposits remain unexcavated, implying that a further three house floors could underlie the current excavation block.The existing radiocarbon data indicate that the site was occupied for 200 -300 years, suggesting that an individual house floor may reflect as little as two decades of deposition.
Both contexts were rich in ecofactual and artefactual remains that could potentially be used for 14 C dating.The most common remains were bone (seal, salmonid, dog, and caribou), seeds (from monocots and edible berries), wood chips, charcoal, insects (fly puparia, lice, and beetles), hair and fur, and grass and wooden artefacts.While many of the ecofacts (such as bone, wood, and charcoal) are frequently preserved on Palaeo-and Neo-Eskimo sites and are commonly used as dating material (e.g., McGhee, 2000;Grønnow and Jensen, 2003), hair, fur, and grass artefacts are extremely rare.More diminutive remains, such as insects and seeds, are probably preserved on many sites and have been recorded in Greenland and the Eastern Arctic (e.g., Zutter, 2009;Dussault et al., 2014;Forbes et al., 2014).However, similar environmental analyses are not routinely conducted in the Western Arctic, and consequently these remains are rarely identified or dated (although see West, 2011).

Radiocarbon Sample Selection for This Study
Myriad processes influence the 14 C age of an ecofact, and clearly not all of the organic remains present at Nunalleq will provide accurate dates (i.e., dates that reflect the true age of the contexts).Six criteria known to influence the accuracy of 14 C ages were used to establish a conceptual hierarchy of suitability for the ecofactual remains present at Nunalleq (Table 2).The marine reservoir effect is by far the most serious problem because it has the potential to skew radiocarbon ages of marine organisms by at least 400 14 C years (e.g., McGhee and Tuck, 1976;Arundale, 1981;Dumond and Griffin, 2002;Walker, 2005).Samples can be of terrestrial origin, but consumption of marine protein (e.g., Betts, 2008), exposure to marine oils (e.g., Morrison, 1989), ingestion of old carbon in the case of insects (e.g., Hoffecker et al., 2012), or consumption of marine plants may result in 14 C ages that do not reflect the assumed isotopic equilibrium with the atmosphere.The potential for a sample to be residual (sensu Bayliss, 2009) is another important factor to consider.For example, curation or reuse of artefacts can result in older materials' being deposited in more recent contexts (e.g., Nelson and McGhee, 2002); TABLE 2. Conceptual hierarchy ranking (from top to bottom) the suitability for 14 C dating of the organic remains present at Nunalleq in relation to the six criteria outlined in Methods.Green shading indicates that the criterion is no cause for concern and orange indicates potential concern, while red shading indicates a high likelihood that the criterion influenced the 14 C measurement.The influence of each criterion as a distorting factor on the 14 C determination is ranked in a perceived decreasing order of severity from left to right.therefore, it is critical that the material being dated reflect the true age of the context (Morrison, 1989;Gerlach and Mason, 1992).A different problem with similar results will occur if organic materials with an inbuilt age (sensu McFadgen, 1982) are dated, resulting in 14 C dates that are centuries older than the deposits from which they were recovered (McGhee, 2009).However, even the selection of short-lived, photosynthetic organisms that should be in isotopic equilibrium with the atmosphere is no guarantee of success, as such samples can be affected by the taphonomic problems discussed above (e.g., West, 2011).Only remains that were widely distributed across the site (and could therefore be routinely used as dating materials) were selected for dating.Eight different ecofacts were selected for analysis.Both seal (Phoca sp.) and salmonid (Oncorhynchus sp.) bone collagen were chosen to establish the extent of the marine reservoir effect and thus establish a baseline for the posited marine mammal oil effect, while wood chips (Picea sp.), being highly absorbent, were selected to evaluate the presence and severity of this effect.Beetle remains (Coleoptera: Olophrum latum Mäklin and Latridius protensicollis Mann., including multiple elytra, heads, and thoraces from each species) and fly (Diptera: Heleomyzidae) puparia were selected to examine the reliability and practicality of 14 C dates on insect remains.The final three ecofacts selected were caribou (Rangifer tarandus) bone collagen, grass (Elymus arenarius) manufactures and seeds from two types of edible berries, cloudberry (Rubus chamaemorus) and crowberry (Empetrum nigrum).Notwithstanding the posited marine mammal oil effect, these samples were anticipated, on the basis of previous studies (e.g., McGhee, 2009), to provide the most reliable ages.

Laboratory Methods and Data Analysis
In total, 16 samples were submitted to the Oxford Radiocarbon Accelerator Unit (ORAU) at the University of Oxford for AMS 14 C dating.In all instances, except for caribou bone and grass samples, the ecofacts selected for dating were isolated from GBA samples collected during excavations in 2014.Subsamples of approximately 100 ml of this material were treated with a weak solution of NaOH (2% -3%) to disaggregate the sample matrix and then sieved using distilled water.The residues were examined under a binocular microscope, and seeds and wood chips were picked for dating.Beetle and fly remains were obtained after processing one-litre subsamples through flotation (Coope and Osborne, 1967;Kenward et al., 1980).These samples were then placed into distilled water with a drop of HCl and stored at 4˚C before being submitted to the ORAU.Grass manufactures and bone were identified in each of the contexts during excavation and cleaned with distilled water before being stored in a dark, dry place.Subsamples were cut from these materials in the Human Palaeoecology Laboratory at the University of Aberdeen before being submitted for dating.
Radiocarbon ages were calibrated using the INTCAL13 calibration curve (Reimer et al., 2013) and a multiplot (Fig. 5) was generated in OxCal (Bronk Ramsey, 2009).If radiocarbon dates on different ecofacts produced overlapping calibrated distributions, their statistical significance was assessed (Fig. 6A) using the Test Sample Significance function in CALIB 7.1 (Stuiver et al., 2016).This method determines a weighted mean of the dates in question and then calculates the test statistic "T" from the weighted sum of the differences between each sample age and weighted mean.The T value is then compared with a chi-square distribution for n − 1 samples.If T exceeds the chi-square value, the dates are deemed to be statistically different ages (Ward and Wilson, 1978).In addition to examining intraecofact statistical equivalences, we conducted an intercontext assessment of the consistency of 14 C ages of all ecofact types (Fig. 6B) to assess the contemporaneity of the deposits.

Context 14035
All of the samples submitted from context 14035 were successfully dated by AMS (Table 3).The youngest measurements, as predicted by the conceptual model (Table 2), were returned on R. tarandus bone collagen (OxA-32659; 211 ± 23) and E. nigrum seeds (OxA-32646; 259 ± 25), which are statistically indistinguishable at 2σ (Fig. 6A).Contrary to the prediction of the conceptual model, the date on E. arenarius (OxA-32958;382 ± 30) was slightly older than those on seeds and caribou bone (Fig. 5).In fact, the measurement on E. arenarius was effectively the same as that on Picea sp. (Fig. 6A) wood chips (OxA-32647;318 ± 24).The wood chips likely derive from the working of driftwood, so their slight antiquity is unsurprising; however, the statistical equivalence with the date on E. arenarius was not anticipated (Table 2) and indicates that the grass manufacture analyzed may be residual to context 14035.The dates on insects were also older than expected.Olophrum latum (OxA-X-2672-12; 440 ± 90) returned an age that was statistically indistinguishable from those on wood and grass, while the 14 C age of Heleomyzidae puparia (OxA-X-2650-27;791 ± 29) was intermediate between the oldest and youngest in the context (Fig. 5).The remaining results followed the predictions of the conceptual model.Seal (Phoca sp.) bone collagen provided the oldest age (OxA-32661; 1005 ± 26), which was slightly older and statistically different from the date on salmonid (Oncorhynchus sp.) bone collagen (OxA-32660;934 ± 23).This marginally younger 14 C age on salmonid bone collagen likely reflects both the anadromous nature and the trophic level of salmon species.Indeed, this hypothesis is borne out in the differences between the stable isotope values of the salmon and seal (Table 4).

Context 14065
Of the eight samples submitted from context 14065, all were successfully dated by AMS except the sample of beetle remains (L.protensicollis), which produced an extremely low carbon yield after pre-treatment.Again, the samples from context 14065 returned results that did not follow the predictions of the conceptual model (Fig. 5).E. arenarius (OxA-32648; 131 ± 24) provided the youngest 14 C age, being both younger than and statistically different (at 2σ) from all of the other samples (Fig. 6A).R. chamaemorus seeds (OxA-32649; 225 ± 31) and R. tarandus bone collagen (OxA-32662; 260 ± 23) were only marginally older than the date on E. arenarius and returned 14 C ages that were indistinguishable from one another (Fig. 6A).As anticipated, the Picea sp.wood chip was slightly older than both caribou bone and cloudberry seeds, which is unsurprising given the absence of trees within the landscape surrounding Nunalleq.The oldest 14 C activity in context 14065, measured on Heleomyzidae (fly) puparia (OxA-X-2650-31; 1016 ± 30), was statistically the same age as the measurement on Phoca sp.bone collagen (OxA-32664; 965 ± 23).As was the case in context 14035, the measurement on Oncorhynchus sp.(OxA-32663; 901 ± 24) returned a slightly younger 14 C activity than that on seal bone, but the dates are statistically equivalent.

DISCUSSION
A necessary first step in assessing the reliability of 14 C measurements is to establish accurately the relative temporal relationships of artefacts and ecofacts from which such measurements were obtained.Once a relative chronology is established, it is possible to make assumptions such as: "A was recovered from a deposit located above B; therefore, the 14 C age of A will likely post-date that of B." If this assumption is disproved, it is then reasonable to seek a reason why the measurement does not conform to expectations.Accurate recording and precise definition of temporal relationships between deposits during excavation are therefore fundamental to procuring "accurate" radiocarbon dates.To achieve this goal, we used open-area excavation and single-context recording to obtain the samples presented here.
Consequently, the samples from this study derive from contexts with a well-defined temporal association (Fig. 4), and we suggest that these contexts represent contemporary episodes of deposition.In order to test this assumption, we examined the statistical consistency of intra-ecofact ages between the two contexts, the hypothesis being that if the contexts are contemporary, their 14 C ages should be statistically indistinguishable.All ecofacts except E. arenarius FIG. 6. Measures of dissimilarity between radiocarbon dates calculated in CALIB using the method of Ward and Wilson (1978).(A) Intra-context comparisons of various ecofact combinations.(B) Inter-context comparisons of the same ecofact type.Where T values exceed the T' 5% value (red dashed line) of 3.84, the two measurements are statistically inconsistent and therefore considered to not represent the same date.manufactures and Heleomyzidae remains (discussed below) returned statistically indistinguishable ages between contexts.These results demonstrate, within the limits of precision afforded by radiocarbon dating, that the seeds, wood chips, and caribou, seal, and fish bone in each context are the same age, or of statistically indistinguishable age.This result strongly supports the temporal associations of the contexts established through excavation and single-context recording.Therefore, having controlled for stratigraphic complexity as a source of unexpected dating results, we can reasonably seek alternative explanations for the divergent results from flies and grass manufactures.In the case of Heleomyzidae, the different 14 C ages may relate to variation within the diet of individual flies, or to differences between species within this fly subfamily.This family comprises species known to breed exclusively on carrion, but also others that breed in mammal and bird excrement or decaying plant matter (Gill and Peterson, 1987;Skidmore, 1995).In context 14065, the age of Heleomyzidae remains was indistinguishable from that of Phoca sp.bone collagen (Fig. 5A), suggesting that perhaps the fossils dated in this instance derive from a species or individuals that fed on seal carrion (cf.Panagiotakopulu et al., 2015).Conversely, the flies dated in context 14035 may have had a different diet, although the stable isotope values (Table 3) are equivocal.
The factors behind the divergent results from artefacts constructed of E. arenarius are less clear, as grass was anticipated to be reliable dating material (e.g., Dumond and Griffin, 2002).In context 14065, E. arenarius returned a 14 C age that was younger than and statistically different from all other ecofacts in the context (Fig. 5, 6).A plausible interpretation could be that E. arenarius reflects the true age of context 14065 and that the dates on R. chamaemorus and R. tarandus are in fact residual.Although tempting, this scenario seems unlikely given the consistency in the two contexts of all of the other measurements (Fig. 6B) and the reliability of dates of caribou bone collagen elsewhere in the Arctic (e.g., Friesen and Arnold, 2008;McGhee, 2000McGhee, , 2009)).More reasonably (given the overlapping calibrated distributions), the divergent dates may reflect the broad plateau associated with the radiocarbon calibration curve, which limits dating precision from the mid-16th century onwards (Burr, 2013), or with measurement uncertainty (Scott, 2013).Alternatively, it is possible that the sample of E. arenarius submitted for dating was contaminated in some way with modern carbon during post-excavation processing, which would have resulted in a younger age (Taylor and Bar-Yosef, 2014).For example, small fragments of more recent plant material could feasibly have become lodged within the weave of the grass manufacture, which would then have influenced the final age of the sample when it was measured in the laboratory.The results from context 14035 present the opposite situation, with E. arenarius being equivalent in age to wood chips of Picea sp. and older than both E. nigrum and R. tarandus.In this instance it would appear that the grass manufacture is residual to the context (sensu Bayliss, 2009).Grass manufactures are typically assumed to have relatively short lifespans of usage, on the order of one to five years.However, this result suggests this may not always be the case.An alternative possibility is that this slightly older datec.120 -170 14 C years older than those on E. nigrum and R. tarandus-is evidence for the marine mammal oil effect cited by a number of researchers (e.g., Park, 1994;Anderson and Freeburg, 2013).Dumond and Griffin (2002) have suggested that the marine reservoir effect in the Bering Sea is on the order of 450 -750 14 C years.This age range is consistent with the dates on seal in this study.In context 14035, the date on seal bone collagen is between 746 and 794 14 C years older than the respective dates on E. nigrum and R. tarandus, while in context 14065, the respective differences between seal and R. tarandus and R. chamaemorus are 705 and 740 14 C years (Table 2).If the slightly older age on E. arenarius were a result of contamination with marine mammal oils, a much greater 14 C age difference would seem likely.Furthermore, the consistency of the 14 C ages of ecofacts between contexts (Figs. 5, 6) and of caribou bone collagen and seed dates, and the absence of unusually old dates on wood, which would presumably be susceptible to the marine mammal oil effect, provide no evidence for this effect at Nunalleq.
In light of the measurements presented here, a reappraisal of the conceptual hierarchy (Table 2) proposed at the outset is necessary.Caribou bone collagen was originally hypothesized to be the most suitable material for dating (c.f.McGhee, 2000;Friesen, 2009), wood was expected to be residual (c.f.Appelt and Gulløv, 2009), and seal and fish bone were anticipated to be the least suitable (c.f.Dumond and Griffin, 2002).The results of this study do not alter these assessments.The main changes relate to recommendations for dating seeds and grass manufactures.As a short-lived terrestrial species, E. arenarius was originally thought likely to be a highly suitable dating material (Table 2).In fact, however, dates on E. arenarius were inconsistent.In context 14065, grass manufactures returned the youngest 14 C age of the study, while in context 14035, their date was older than expected.The first of these results suggests that grass manufactures may be susceptible to contamination, while the second indicates that although E. arenarius is short-lived, the artefacts constructed from it may have longer use periods than previously assumed.As for seeds, those from edible berries are abundant in the archaeological remains of sod structures since they are part of the fabric of the peat blocks used to construct such dwellings.Consequently, the potential for seeds of edible berries to be residual to archaeological contexts was judged to be high, and they were ranked less suitable for dating than caribou bone collagen.However, the consistency of the 14 C ages on seeds between the two contexts, and between caribou, crowberry, and cloudberry within the same context (Fig. 6A and B), provides no evidence for the seeds used for this study having been residual.Therefore, on the basis of this evidence, edible berry seeds can be said to be as reliable as caribou bone collagen as a dating material.

CONCLUSIONS
Recent decades have witnessed a gradual reassessment and greater appreciation of the complexities of radiocarbon data by archaeologists working in the North American Arctic.Few researchers continue to procure 14 C dates on sea mammal bones and wooden artefacts, and there is a growing acceptance that terrestrial macrofossils are the key to accurate radiocarbon chronologies.Despite these advances, problems remain.For instance, there is often only cursory recognition that accurate relative chronology (for sites and dating materials) is essential if radiocarbon data from the Arctic, and their inherent taphonomic issues, are to be fully understood.Indeed, despite frequent calls, there has been little engagement with assessing the respective 14 C dating suitability of different organic remains.
This study addressed both of these issues.Using openarea excavation and single-context recording, we ensured a robust interpretation of the temporal relationships of the materials analyzed.Statistical equivalence of intraecofact 14 C measurements between the two contexts confirmed these measurements and provided us with greater confidence that further 14 C age variation was related to taphonomy and not stratigraphy.These results demonstrate that caribou bone collagen is a reliable and accurate dating material, but more importantly, they indicate that seeds from edible berries (Rubus chamaemorus and Empetrum nigrum) also produce reliable dates that are statistically indistinguishable from those on caribou.Contrary to expectations, grass manufactures were unreliable, returning different ages in the two contexts.In context 14065, they were younger than caribou bone and seeds, while in context 14035, they were older.Measurements on other materials either confirmed well-established trends or followed expectations.
Despite the successes of this study, it is important to recognize its limitations.Only two contexts were analyzed, and to draw more statistically significant conclusions, a larger dataset covering more contexts, but testing fewer ecofact types (i.e., seeds, caribou bone, seal bone, wood, and grass) would be appropriate.An example would be demonstrating statistical equivalence of 14 C dates on seeds and caribou across several stratigraphically contemporary archaeological contexts.Furthermore, it is essential to stress the applicability of these findings.The ecofacts dated in this study derive from clearly delineated house floors, and therefore they could reasonably be assumed to have been deposited during the formation of these archaeological layers.Although ecofacts such as seeds are likely to be abundant in archaeological deposits from the Arctic, there is little value dating seeds of uncertain provenance.Such an ecofact could easily derive from the peat used to construct the dwelling and would therefore not be dating the archaeological event of interest.
Context, in the sense of understanding the relative relationship between archaeological deposits, is therefore the key to improved dating.Our own experiences at Nunalleq, where we transitioned from planum to singlecontext recording, support this.The original suite of dates, on material sourced from multiple squares, provided a coarse estimate for the duration of occupation at the site, but with some inconsistencies and a limited means for assessing the reliability of these measurements.In contrast, single-context recording has provided a relative chronology for all of the archaeological deposits at Nunalleq, while also conferring confidence in the results of this study.Going forward with this robust relative chronology, it will be possible to use Bayesian statistics to construct highly resolved (sub-centennial) archaeological datasets to tackle questions pertaining to dietary and technological change, shifting subsistence patterns, and responses to subcentennial climatic variation.
FIG. 4. (A)Photograph (facing southwest) of the excavation at Nunalleq following the removal of debris deposits(contexts 14006, 14015, 14021, 14007, 14034,  14033, 13012, 14048, 14049, 14050)  associated with the collapse of the dwelling.The white line reflects the line of the cross-section presented in C, and the dotted yellow lines illustrate the extent of contexts 14035 and 14065.Grid squares (2 × 2 m) and their associated numbers are indicated in red.(B) Partial Harris matrix illustrating the stratigraphic relationships of the deposits intersected by the cross-section presented in C. Ovals represent deposits or fills and squares indicate cut features.Deposits are colour coded: brown = debris, red = house floors, purple = fills, blue = construction or remodeling deposits, and orange = structural deposits.(C) Reconstructed cross-section of the latest phase of activity constructed using the contextual information (including levels and spatial extent) recorded during excavation (c.f.Brown and Harris, 1993).

TABLE 1 .
Radiocarbon rangefinder dates from samples excavated during the rescue excavations in 2009 and 2010.NR = not recorded.
FIG. 2. Rangefinder radiocarbon dates undertaken on materials excavated using planum methods in 2009 to 2012.FIG. 3. Schematic site plan indicating the locations of the contexts examined in this study.

TABLE 3 .
Radiocarbon dates from this study.

TABLE 4 .
Stable isotope values of selected ecofacts.