INTRODUCTION
The North Atlantic region has a very rich archaeological and paleoecological record in which marine resources feature prominently; these sample types are almost ubiquitous in, for example, coastal middens, where they can be essential materials for radiocarbon dating. Marine artifacts and ecofacts on archaeological sites arise via significant use and consumption of marine resources at particular time periods, by prehistoric and historic communities across the North Atlantic. Specific examples include Mesolithic societies along the Atlantic coast of Europe (Noe-Nygaard Reference Noe-Nygaard1988; Lubell et al. Reference Lubell, Jackes, Schwarcz, Knyf and Meiklejohn1994; Richards and Hedges Reference Richards and Hedges1999), at the Mesolithic–Neolithic transition in the UK (Schulting and Richards Reference Schulting and Richards2002; Montgomery et al. Reference Montgomery, Beaumont, Jay, Keefe, Gledhill, Cook, Dockrill and Melton2013), and during the Viking period on the North Atlantic islands (including Scotland, Faroes, Iceland, and Greenland; Arneborg et al. Reference Arneborg, Heinemeier, Lynnerup, Nielsen, Rud and Sveinbjorndóttir1999, Reference Arneborg, Lynnerup and Heinemeier2012; Barrett et al. Reference Barrett, Beukens and Nicholson2001; Ascough et al. Reference Ascough, Cook, Church, Dugmore, Arge and McGovern2006, Reference Ascough, Church, Cook, Dunbar, Gestsdóttir, McGovern, Dugmore, Friðriksson and Edwards2012). Stable isotopes of carbon and nitrogen (δ13C and δ15N) have been used in these studies to demonstrate the incorporation of significant amounts of marine material in human diets, according well with the archaeology of these time periods, in which fishing vessels, equipment for fish and shellfish collection and processing, and other material remains of these activities are found. One location in which marine resources were used almost continuously during the Holocene is Scotland, with its extensive coastline and island archipelagos to the west and north. Scottish archaeology represents a very detailed record of North Atlantic communities over the past ~10,000 yr and is important for its position at the interface between Europe and the North Atlantic region, making it key to our understanding of factors such as cultural adaptation to climatic and environmental changes in marginal environments, human-environment interactions, and trade and exchange over extended distances.
In order to understand the chronology of events in Scottish archaeology, 14C dating is crucial for building absolute chronologies within the archaeological and paleoenvironmental sciences. However, the use of marine resources introduces the need to correct 14C dates for the marine reservoir effect. A reservoir effect occurs when the carbon within one of Earth’s carbon reservoirs (i.e. the terrestrial biosphere, marine or freshwater hydrospheres, or the cryosphere) has a lower 14C activity (and hence an older “apparent” 14C age) than carbon in the atmosphere. This can occur if ancient carbon (e.g. carbon from carbonate rocks such as limestone) enters the reservoir, or if carbon undergoes “aging” within the reservoir as a result of time spent in that reservoir without exchange. As global circulation of 14CO2 in the atmosphere is rapid, being on the order of 5–10 yr (Levin and Hesshaimer Reference Levin and Hesshaimer2000), and uptake of 14CO2 by plants and subsequent transfer through the food chain is equally rapid (Nydal Reference Nydal1968), terrestrial environments do not typically have a reservoir effect, with the exception of material in close proximity (<1 km) to volcanic CO2 sources (Bruns et al. Reference Bruns, Levin, Münich, Huberten and Filipakis1980). In contrast, the marine reservoir exhibits a substantial 14C reservoir effect due to the “aging” of deep water masses when separated from the atmosphere. When these water masses return to the surface, they “dilute” the 14C content of the surface ocean, and this dilution is passed to organisms inhabiting the marine reservoir (e.g. fish and mollusks). Importantly, the reservoir effect is also transferred to terrestrial organisms, such as humans, that consume marine resources.
Therefore, 14C-dated remains of marine material from archaeological sites require correction for the marine reservoir effect (MRE), as do the remains of humans and other omnivores that have demonstrably consumed a significant proportion of marine carbon in their diet. Without correction, samples can appear several hundred years “too old,” leading to incorrect chronologies of events in the archaeological record. For example, uncorrected dates on marine material from wheelhouse sites on the Western Isles of Scotland appear to show that these structures are equivalent in age to the demonstrably earlier architectural form of brochs, yet when corrected for the MRE this discrepancy is removed (Barber Reference Barber2003; Ascough et al. Reference Ascough, Cook, Dugmore, Barber, Higney and Scott2004). Clearly, in order for the resulting 14C dates to be accurate, the MRE correction needs to be appropriate to the individual site, period, and samples. The marine calibration curve (currently Marine13; Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013) gives a global average MRE correction that varies with time. However, individual locations around the globe are offset from this average value, where the offset is known as ΔR (Stuiver et al. Reference Stuiver, Pearson and Braziunas1986; Stuiver and Braziunas Reference Stuiver and Braziunas1993). These ΔR values are location-specific and can vary at a single location through time, making their quantification an important issue for 14C dating of marine material. Spatiotemporal variability in MRE and ΔR values can result from several different oceanographic, environmental, or climatic factors. These include changes in ocean circulation that bring water masses of varying 14C content to an area, changes in ocean ventilation or stratification that increase or reduce the input of 14C-depleted waters from depth, fluctuations in wind speed, air/water temperature or ice cover affecting ocean uptake of atmospheric 14C, and in estuarine settings, changes in the admixture of fresh (high 14C) and marine (low 14C) waters.
The North Atlantic has been the setting of extensive efforts to quantify regional MRE and ΔR values. Modern ΔR values range from 225±51 14C yr in Kollafijord, Iceland (Broecker and Olson Reference Broecker and Olson1961), to –119±54 14C yr at Skelmorlie Bank, Scotland (Harkness Reference Harkness1983), with a clear geographic gradient from Arctic waters containing proportionally “older” carbon (i.e. high MRE) in the north, to Atlantic waters with lower MRE values further south, a trend that appears to have been in existence through at least the last 1000 yr (Ascough et al. Reference Ascough, Cook, Church, Dugmore, Arge and McGovern2006). In waters surrounding Scotland, modern values of ΔR range from –119±54 to +94±30 14C yr (Harkness Reference Harkness1983), while nonmodern values have been measured at –123±62 to +143±20 14C yr (Ascough et al. Reference Ascough, Cook, Dugmore and Scott2007; Russell et al. Reference Russell, Cook, Ascough and Scott2015) during the Holocene. In the pre-Holocene North Atlantic, there are large shifts to higher MRE values on the order of several hundred years during the Younger Dryas (i.e. ~11,000 yr BP; Austin et al. Reference Austin, Bard, Hunt, Kroon and Peacock1995), and shifts on the order of 1000 yr during the Last Glacial (Skinner and Shackleton Reference Skinner and Shackleton2004). Recent work by Russell et al. (Reference Russell, Cook, Ascough and Scott2015) failed to detect significant shifts in ΔR in Scottish coastal waters over the latter half of the Holocene, although five outliers from this trend were detected. This work was based upon multiple paired sampling and a statistical approach that involves “bootstrapping” to determine the likelihood that repeat measurements would give the same ΔR for a location if different samples were selected. The approach involves taking multiple samples of terrestrial and marine material, and for every possible terrestrial-marine sample pairing, calculating a ΔR value. The weighted mean of these values is taken as the overall ΔR for a context, and the uncertainty on this weighted mean is obtained by combining the standard error of the weighted mean with the standard deviation of all calculated ΔR values (i.e. the standard error for predicted values). In conclusion, Russell et al. (Reference Russell, Cook, Ascough and Scott2015) recommended using a ΔR value of –47±52 14C yr for the period 3500 BC to AD 1450 in Scottish coastal environments if no further information for a specific site and time period is available. This value overlaps with a previous determination for the subpolar eastern North Atlantic (including Scotland and Ireland), for the mid- to late Holocene by Reimer et al. (Reference Reimer, McCormac, Moore, McCormick and Murray2002), which was –33±93 14C yr.
The five outliers from the Russell et al. (Reference Russell, Cook, Ascough and Scott2015) data set include material from the Neolithic period (Carding Mill Bay, 3640–3520 BC) and the Medieval period (Roberts Haven, AD 1280–1390), both of which are critical for understanding the chronology of Scotland’s archaeology. The Scottish Mesolithic/Neolithic transition ~6000 cal BP saw the introduction of organized farming practice for the first time, while the Medieval period saw the expansion of trade routes with Europe and further afield, based upon an emergent fishing industry (Barrett et al. Reference Barrett, Johnstone, Harland, Van Neer, Ervynck, Makowiecki, Heinrich, Hufthammer, Enghoff, Amundsen, Christiansen, Jones, Locker, Hamilton-Dyer, Jonsson, Lõugas, Roberts and Richards2008). We therefore sought further information on the MRE in Scotland for these time periods by 14C analysis of paired marine and terrestrial samples from four archaeological sites. We also recalculated ΔR values for two further sites that were not contained within the Russell et al. (Reference Russell, Cook, Ascough and Scott2015) paper, but which relate to the Mesolithic periods in Scotland. The aim of this work was therefore to clarify MRE values for important periods in Scottish prehistory to improve archaeological chronologies. In addition, we examined the effect of taphonomic bias upon MRE values, and critically assessed the multiple paired sample approach for MRE and ΔR quantification.
METHODS
Sample Selection
Samples were selected for new quantifications of the MRE from individual stratigraphic contexts at four sites: Context 14 at Northton (NO-14); Context 1 at Tràigh na Beirigh 1 (TNB1-1); Context 5 at Tràigh na Beirigh 2 (TNB2-5); and Context 177/83 at Guinnerso (GUN-177/83). Northton (NGR: NF 9753 9123) is located on the Isle of Harris, Scotland, while Tràigh na Beirigh 1 & 2 (NGR: NB 1002 3628 & NB 1003 3633) and Guinnerso (NGR: NB 0350 3631) are located on the Isle of Lewis, Scotland (Figure 1). The archaeological evidence at Northton consists of a series of Mesolithic ground surfaces with mixed anthropogenic material within the soils that is overlain by machair, a calcareous shell-sand soil unique to the Western Isles of Scotland. The site represents the first archaeological evidence for Mesolithic human occupation in the Western Isles (Gregory et al. Reference Gregory, Murphy, Church, Edwards, Guttmann and Simpson2005) and the samples for this project were taken from the latest Mesolithic layer, immediately under the machair (Bishop et al. Reference Bishop, Church and Rowley-Conwy2010). The sites of Tràigh na Beirigh 1 & 2 consist of two open-air Mesolithic shell middens, again overlain by machair (Church et al. Reference Church, Bishop, Blake, Nesbitt, Perri, Piper, Rowley-Conwy, Snape-Kennedy and Walker2012; Bishop et al. Reference Bishop, Church, Clegg, Johnson, Piper, Rowley-Conwy and Snape-Kennedy2013). The samples for this project (TNB1-1, TNB2-5) were taken from the main body of the shell middens at both sites. The final samples (GUN 177/83) come from the Medieval occupation of a shieling (a stone and turf hut forming a summer dwelling in a seasonal upland pasture or heathland) located in the multiperiod landscape at Guinnerso in the moorland of the Uig Peninsula in Lewis (Church and Gilmour Reference Church and Gilmour1998).
Location of sample sites from which material was obtained for MRE/ΔR quantification, from which data was recalculated, and locations mentioned in the text (SA=Sand; CMB=Carding Mill Bay; NO=Northton; TNB=Tràigh na Beirigh; GUN=Guinnerso).
The selected sites are exposed to the Atlantic Ocean, away from significant sources of freshwater or carbonate geology, either of which could compromise 14C dates used to quantify the MRE. Selection of contexts followed the processes described in Ascough et al. (Reference Ascough, Cook, Dugmore, Scott and Freeman2005). Briefly, material was only selected from discreet, sealed contexts of limited spatial extent, without visible signs of disturbance. These sites were selected after an initial program of range-finder 14C dating sponsored by Historic Environment Scotland indicated that NO-14 corresponded to the mid-Mesolithic period, TNB1-1 and TNB2-5 to the latest Mesolithic period, and GUN-177/83 to the Medieval period. At each site, four paired samples of terrestrial material (carbonized plant macrofossils) and marine material (marine mollusk shells) were selected from bulk samples taken for environmental archaeological analysis, using an on-site “total” sampling strategy, following Jones (Reference Jones1991). Bulk samples were processed using a flotation tank (Kenward et al. Reference Kenward, Hall and Jones1980), with the residue held by a 1.0-mm net and the flot caught by 1.0- and 0.3-mm sieves, respectively. All the flots and residues were air-dried and sorted using a low-powered stereo/binocular microscope at 15× to 80× magnification. Hazel nutshell (Corylus avellana L.) were chosen as the terrestrial single-entity samples from the Mesolithic sites, as hazelnuts are short-lived, single-season plant remains and are very common on Mesolithic sites in Scotland (Bishop et al. Reference Bishop, Church and Rowley-Conwy2014, Reference Bishop, Church and Rowley-Conwy2015). Barley grains were chosen from the Medieval phase at Guinnerso as they too are short-lived, single-season plant remains. Common limpet shells (Patella vulgata L.) were selected from all four sites as the marine sample to which the 14C ages of the hazelnut and barley (terrestrial) samples were compared. The lifespan of the common limpet ranges from ~5 to ~20 yr (Lewis and Bowman Reference Lewis and Bowman1975), introducing the possibility of inbuilt ages of up to 20 yr when using limpets to calculate MRE and ∆R. In this study, shells were inspected to estimate age based upon growth bands where possible, and to select shells <10 yr old. Shell morphology was also checked to ensure this was consistent with the faster-growing, shorter-lived individuals at the lower shoreline (Lewis and Bowman Reference Lewis and Bowman1975). Any inbuilt age associated with marine shells used in this study will therefore be low, compared to the typical uncertainties associated with MRE and ∆R determinations. Although species-specific MRE and ∆R values for marine mollusk shells have been observed at locations worldwide, these are highly unlikely for the study region. Species-specific effects arise where there are differences in 14C age of resources consumed by mollusks, typically in areas of carbonaceous geology where infaunal feeders will ingest 14C-dead carbon during feeding (cf. Forman and Polyak Reference Forman and Polyak1997). Species-specific effects can also arise where there are significant differences in 14C age of the water column over small geographical areas, such as estuaries (e.g. Holmquist et al. Reference Holmquist, Reynolds, Brown, Southon, Simms and MacDonald2015). Neither of these applies in the study area, and previous work has showed no interspecies variability in mollusk MRE and ∆R values for the region (Ascough et al. Reference Ascough, Cook, Dugmore, Scott and Freeman2005).
In addition to new MRE quantification, recalculations of the MRE and ∆R were performed for two sites relating to the early Holocene and Mesolithic period in Scotland; Northton on the Isle of Harris (context NO-5) and Sand on the Scottish mainland (context SA-13) (Figure 2) (Ascough et al. Reference Ascough, Cook, Dugmore and Scott2007), in order to assess these data in light of the findings presented in Russell et al. (Reference Russell, Cook, Ascough and Scott2015). ∆R values and terrestrial calibrated age ranges for these sites were therefore recalculated using the IntCal13 and Marine13 data sets (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013), and the standard error for predicted values, outlined in Russell et al. (Reference Russell, Cook, Ascough, Scott and Dugmore2011a, Reference Russell, Cook, Ascough, Barrett and Dugmore2011b, Reference Russell, Cook, Ascough and Scott2015), was calculated for each ΔR value obtained. This is particularly important for these two sites as they previously gave ∆R values of 64±19 14C yr (SA-13) and 79±32 14C yr (NO-5) (Ascough et al. Reference Ascough, Cook, Dugmore and Scott2007). These data were taken to indicate that ∆R values were higher in the early Holocene/Mesolithic as they related to the periods 6480–6420 BC (SA-13) and 6390–6230 BC (NO-5) (Ascough et al. Reference Ascough, Cook, Dugmore and Scott2007). By recalculating these data using the standard error for predicted values (Russell et al. Reference Russell, Cook, Ascough and Scott2015), we can assess whether this more robust method of estimating the error on ∆R values still gives values for Scottish waters that are significantly different from those later in the Holocene period.
Graph of ΔR values for Scottish coastal waters through the Holocene showing new values (black squares) and recalculated values (gray triangles) alongside previous values for Scottish waters (white circles: Ascough et al. Reference Ascough, Cook, Dugmore, Barber, Higney and Scott2004, Reference Ascough, Cook, Church, Dugmore, Arge and McGovern2006, Reference Ascough, Cook, Dugmore and Scott2007, Reference Ascough, Cook and Dugmore2009; Russell et al. Reference Russell, Cook, Ascough and Dugmore2010, Reference Russell, Cook, Ascough, Barrett and Dugmore2011b, Reference Russell, Cook, Ascough and Scott2015).
Radiocarbon Measurement of Samples for MRE/ΔR Quantification
Carbonized plant macrofossils were pretreated with a HCl wash to remove carbonates (0.1M at 80°C for 2 hr), followed by removal of organic acids in 0.1M NaOH (2 hr at 80°C), then a final HCl wash to remove any CO2 adsorbed in the base step. The pretreated macrofossils were converted to CO2 by combustion in precleaned quartz tubes (Vandeputte et al. Reference Vandeputte, Moens and Dams1996). Marine shells were inspected to establish that there was no evidence of carbonate reprecipitation (Mangerud Reference Mangerud1972; Mook and Waterbolk Reference Mook and Waterbolk1985). Shells were cleaned ultrasonically and by abrasion to remove surface contaminants, and then etched in 1M HCl to remove the outer 20% of the shell. The whole shell was then crushed and a 0.1-g aliquot was hydrolyzed with 1M HCl under vacuum. CO2 from plant or shell samples was purified cryogenically using solid CO2/ethanol and liquid N2 traps. Aliquots of 3 mL purified CO2 were converted to graphite by the method of Slota et al. (Reference Slota, Jull, Linick and Toolin1987), and sample 14C/13C ratios were measured by accelerator mass spectrometry (AMS). δ13C values (as per mil, ‰, deviations from the VPDB international standard) were measured on CO2 from all samples using a VG SIRA 10 with NBS 22 (oil) and 19 (marble) as internal standards. The full methodology is given in Dunbar et al. (Reference Dunbar, Cook, Naysmith, Tripney and Xu2016).
Consistency of 14C Measurements within Sample Groups
The groups of measured terrestrial and marine 14C ages for the individual contexts were tested for internal consistency using the chi-squared (χ2) test (cf. Ward and Wilson Reference Ward and Wilson1978). The test establishes whether a group of 14C ages can be considered to be contemporaneous by comparing the variability within a measurement group with the errors on individual measurements. Measurement variability is considered to exceed that occurring by chance (i.e. χ2 test fail) if the χ2 test value (T) for a group of 14C ages exceeds the T statistic for 95% confidence of n 14C age measurements (χ2 0.05 T). If a group of samples failed the χ2 test, the measurements were scrutinized to establish the source of the variation. Where the χ2 test fail was due to a single measurement, this measurement was excluded from the sample group, and the remaining consistent 14C measurements used to calculate ∆R. In instances where the χ2 test fail was due to multiple measurements, the 14C dating of the context was repeated where possible, using additional samples (cf. Ascough et al. Reference Ascough, Cook, Dugmore and Scott2007).
Calculation of ΔR Values
For each context, multiple values of ∆R were calculated using samples that passed the χ2 tests. Two slightly different methods of calculating ∆R exist; therefore, we performed a sensitivity test, comparing the results obtained with both methods to check for any significant differences. The first method involves converting individual terrestrial 14C ages to modeled marine 14C ages using an interpolation of the IntCal13 and Marine13 data sets (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013). The conversions incorporate the uncertainty in the interpolated calibration curve data. The ∆R for each pairing of terrestrial and marine 14C ages is the difference between the midpoint of the modeled marine 14C age boundaries and the measured marine 14C age. The 1σ error on individual ∆R values was calculated by propagation of the errors on the terrestrial and marine 14C ages.
The second method differs slightly in that it incorporates the probability density function (PDF) of the marine calibration curve when obtaining ∆R (Reimer and Reimer, forthcoming). The individual terrestrial 14C ages are calibrated using the IntCal13 calibration curve. This produces a PDF, the discreet points of which are reverse-calibrated using the marine calibration curve. The offset between the 14C-dated marine sample and the reverse-calibrated terrestrial sample PDF gives ∆R. To determine the confidence interval of ∆R, a convolution integral is used, approximated as a normal distribution (Reimer and Reimer Reference Reimer and Reimer2016).
For both methods, ∆R was calculated for each possible pairing of marine and terrestrial 14C measurements for a context, giving multiple ∆R values for that context. The weighted mean of the ∆R values was then calculated to give an overall ∆R value for that context. The standard error on the weighted mean was evaluated based upon the measurement uncertainties (Equation 1):
$$\sigma _{1} {\equals}\sqrt {{1 \over {\mathop{\sum}{{1 \mathord{\left/ {\vphantom {1 {s_{i}^{2} }}} \right. \kern-\nulldelimiterspace} {s_{i}^{2} }}} }}} $$
The final 1σ error associated with a weighted mean ∆R for a context was then obtained via the standard error for predicted values. This accounts for any additional variability due to the precise pairing of terrestrial and marine samples used to calculate ∆R
$$\sigma {\equals}\sqrt\left( {x^{2} {\plus}y^{2} } \right)$$
where x=error on the weighted mean and y=standard deviation on all the ∆R values calculated for a context.
Terrestrial Calibrated Age Ranges
To calculate a calendar age range that is represented by the material in the deposit (and for which the ΔR values are applicable), the weighted mean of the terrestrial 14C ages that passed the χ2 test for each context was used. Calibrated ranges at 95% confidence (i.e. 2σ) were obtained using the IntCal13 atmospheric data set (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013), and the OxCal v 4.2 calibration program (Bronk Ramsey Reference Bronk Ramsey1995, Reference Bronk Ramsey2001).
RESULTS
New ΔR Values for the Mesolithic and Medieval Periods
The δ13C values for carbonized plant macrofossils and marine shells fall within the expected ranges for these sample types (i.e. C3 vegetation in the Northern Hemisphere and marine carbonates; Aitken Reference Aitken1990). The χ2 test results for the groups of terrestrial and marine samples for each context are given in Table 1, along with the 14C ages and δ13C values for each sample. The reported χ2 test results are for groups of samples where the variability in 14C measurements did not exceed the T value, and results were used in assessment of MRE/ΔR for that context. Samples that caused the 14C measurements in a group to fail the χ2 test are indicated; these measurements were excluded from ΔR calculation.
Results of δ13C values, 14C measurements±1σ, and χ2 test results for samples measured in this study.
* Measurements excluded from the sample group on the basis of the χ2 test.
The sensitivity test between the two methods of calculating ΔR showed no significant differences, with a maximum of 12 14C yr between ΔR values calculated using different methods. The confidence interval of the PDF method is, however, slightly larger than that obtained using the intercept method; therefore, we use the PDF method to report ΔR values in the following, as the results are the more conservative of the two methods.
For TNB1-1, the weighted mean 14C age of the terrestrial samples gives a calibrated age range of 4330–4240 BC (6280–6190 BP) at 95% confidence, placing this site at the latest phase of the Mesolithic in Scotland (Ashmore Reference Ashmore2004). For this time period, the calculated MRE is 300±51 14C yr and the ΔR=–109±55 14C yr. This ΔR value overlaps with that calculated for TNB2-5, which is –143±54 14C yr, corresponding to a MRE value of 229±41 14C yr for the period 4540–4470 BC (6490–6410 BP), in the late Scottish Mesolithic (Ashmore Reference Ashmore2004). For GUN-177/83, the weighted mean terrestrial 14C age corresponds to the Late Medieval period, at AD 1460–1630 (490–320 BP). For this time interval, the calculated MRE is 305±24 14C yr and the ΔR=–130±36 14C yr. The group of terrestrial 14C ages for NO-14 are statistically consistent, with a χ2 value for the group of T=2.13 (χ2 0.05=7.81), giving a weighted mean age of 7450±17 14C yr BP, and a calibrated age range of 6390–6250 BC (8330–8200 BP). The group of marine ages are also internally consistent, with a χ2 value for the group of T=0.53 (χ2 0.05=7.81). The weighted mean of the terrestrial group of samples is 2361 14C yr older than the group of marine samples. When the MRE is responsible for an age offset between terrestrial and marine samples, the marine material is always older than the terrestrial samples. As the reverse is true in this instance, the age offset between terrestrial and marine samples for NO-14 must be that the two sample types are of different actual ages, and entered the context ~2000 14C yr apart. It is therefore not possible to calculate a MRE for NO-14 using these samples.
Recalculation of ΔR Values Using the Standard Error for Predicted Values
Recalculated ΔR and MRE values with the standard error for predicted values for SA-13 and NO-5 are given in Table 2. For SA-13, this gives a MRE of 416±35 14C yr and a ΔR of 63±49 14C yr for the period 6480–6420 cal BC (8430–8370 cal BP). For NO-5, this gives a MRE of 440±69 14C yr and a ΔR of 67±78 14C yr for the period 6390–6290 cal BC (8340–8180 cal BP).
MRE values, ΔR values, and calibrated terrestrial calendar age ranges (95% confidence interval) for samples analyzed in this study.
* Values not calculated due to taphonomic disturbance.
DISCUSSION
Archaeological Significance of the Dating Program
The terrestrial dates from the hazel nutshell and barley carbonized macrofossils from the four sites are important in determining the chronology of the sites excavated. The four hazel nutshell dates from Northton (NO-14) have demonstrated that the latest paleosol in the site sequence is of the same date as the main Mesolithic archaeological phase at the site dating to the 7th millennium BC (Gregory et al. Reference Gregory, Murphy, Church, Edwards, Guttmann and Simpson2005), albeit with some later intrusion from the later Neolithic archaeology in the machair overlying the paleosol sequence. The hazel nutshell dates from the two open-air shell middens at Tràigh an Beirigh (TNB1-1 & TNB2-5) date the activity at these sites to the 5th millennium BC, furnishing the archaeology of the Western Isles of Scotland with Terminal Mesolithic shell-midden sites for the first time. These open-air shell middens are one of the main site types of the Late Mesolithic in Scotland and the wider European Atlantic seaboard (Milner et al. Reference Milner, Craig and Bailey2007; Hardy Reference Hardy2015) and the lack of these sites in the Western Isles until this point has been viewed as an enigmatic problem in North Atlantic archaeology (Edwards Reference Edwards1996; Hardy Reference Hardy2015). The barley dates from the Medieval shieling at Guinnerso (GUN-177/83) also demonstrate the antiquity and importance of transhumance practice in the Western Isles.
New Determinations of MRE and ΔR Values for the Mesolithic and Medieval Periods in the Western Isles of Scotland
The results of this study fill important gaps in our knowledge of 14C dynamics in ocean systems surrounding Scotland that relate to the 14C dating of historic and prehistoric communities in the region. For the earliest period covered, the recalculated values for SA-13 and NO-5 relate to the periods 6480–6420 and 6390–6290 cal BC, respectively, corresponding to the Mesolithic period. There is only a 29-yr gap between the two calibrated age ranges, meaning that the ΔR values from these sites both correspond to one of the earliest periods represented in Scottish archaeology. The ΔR values for SA-13 and NO-5 are statistically indistinguishable on the basis of a χ2 test, and can be combined to give a weighted mean of 64±41 14C yr. The recalculated ΔR values are equivalent (on the basis of a χ2 test with df=38) to 37 other values for Scottish coastal waters in the Holocene, presented in Russell et al. (Reference Russell, Cook, Ascough and Scott2015). However, as the results from SA-13 and NO-5 give slightly higher MRE/ΔR values for the earliest period covered, possible factors underlying this can be considered. One possibility is that the age ranges for SA-13 and NO-5 follow the 8200-yr event in paleoenvironmental records for the North Atlantic region (Alley and Ágústsdóttir Reference Alley and Ágústsdóttir2005). A proposed mechanism for this event is the catastrophic drainage of two large glacial lakes, Agassiz and Ojibway, into the North Atlantic (Barber et al. Reference Barber, Dyke, Hillaire-Marcel, Jennings, Andrews, Kerwin, Bilodeau, McNeely, Southon, Morehead and Gagnon1999). This influx of freshwater may have resulted in a slowdown of the North Atlantic Deepwater (NADW) Conveyor, consequently resulting in colder conditions in the region (Ellison et al. Reference Ellison, Chapman and Hall2006). A NADW slowdown period is thought to be followed by phases of “older” surface ocean ages, as “aged” deep waters are returned to the surface (Thiagarajan et al. Reference Thiagarajan, Subhas, Southon, Eiler and Adkins2014). Regardless of the mechanism for change in MRE/ΔR, the use of values from NO-5 and SA-13 are recommended for this time period until further data become available.
For the latest Mesolithic period, values from TNB1-1 and TNB2-5 are statistically equivalent, with a χ2 value of T=0.19 (χ2 0.05=3.84). This indicates a ΔR of –126±39 14C yr for the Western Isles of Scotland during the period 4540–4240 BC. It is important to note that there is a 135-cal yr hiatus between the upper and lower limits of the two calibrated ranges making up this timespan. Both TNB1-1 and TNB2-5 are statistically indistinguishable from (on the basis of a χ2 test with df=37) the values presented in Russell et al. (Reference Russell, Cook, Ascough and Scott2015). The closest ΔR values in time for this geographic region are obtained from Carding Mill Bay (CMB), which has a lower calibrated age limit of 3641 BC, putting a gap of 596 cal yr between this and the upper limit of TNB1-1. The ΔR for CMB was an outlier from other Holocene values in Russell et al. (Reference Russell, Cook, Ascough and Scott2015), being significantly higher (ΔR=150±28 14C yr for the period 3641–3521 BC). Two previous values for CMB in Reimer et al. (Reference Reimer, McCormac, Moore, McCormick and Murray2002) give a ΔR of –44±91 14C yr for the period 3965–3714 BC and ΔR=86±67 14C yr for 3942–3653 BC. The spread in these determinations is large, although the calibrated age range for the positive ΔR obtained in Reimer et al. (Reference Reimer, McCormac, Moore, McCormick and Murray2002) is closest in time to the highly positive ΔR presented in Russell et al. (Reference Russell, Cook, Ascough and Scott2015). It is possible that in coastal waters surrounding CMB there were significant fluctuations in ΔR over the time period 3965–3521 BC. Potential mechanisms for these fluctuations include varying proportions of high 14C content Atlantic water reaching the site through time due to oceanographic shifts. If this were the case, other sites in the region would also be expected to show concurrent ΔR changes; however, we currently lack these measurements. Overall, a reassessment of data from CMB would be useful in light of these data. For the latest Mesolithic period in the Western Isles of Scotland, we therefore recommend using the ΔR values calculated for TNB-1-1 and TNB2-5. This correction would be applicable to marine samples that return 14C ages around 5908±21 to 5697±21 yr BP (the weighted means of marine 14C ages for TNB-1-1 and TNB2-5, respectively). Prior to these new ΔR calculations, there was a gap of 2647 calendar years for which no values were available. Determinations of accurate MRE/ΔR values for this period are especially important given the evidence for marine consumption during the Mesolithic in Scotland (Schulting and Richards Reference Schulting and Richards2002) and the debate surrounding whether the use of marine resources continued into the Neolithic (Milner et al. Reference Milner, Craig, Bailey, Pedersen and Andersen2004; Montgomery et al. Reference Montgomery, Beaumont, Jay, Keefe, Gledhill, Cook, Dockrill and Melton2013).
Previous data for the Medieval period suggested a slightly elevated ΔR relative to the preceding Norse period (Ascough et al. Reference Ascough, Cook and Dugmore2009). However, when the standard error for predicted values was applied, these values were not found to be significantly different from other values for Scottish waters during the period 3500 BC–AD 1450 that were used to calculate an average ΔR for this time period of –47±52 14C yr (Russell et al. Reference Russell, Cook, Ascough and Scott2015). The exception to this was values of ΔR for Atlantic cod (Gadus morhua L.) at Roberts Haven (AD 1284–1393), which were higher (105±34 14C yr) (Russell et al. Reference Russell, Cook, Ascough, Barrett and Dugmore2011b), which may indicate integration of ΔR values over a wider geographic range, including northern waters, where higher ΔR values are found. For the time period of AD 1457–1632, the ΔR value at GUN-177/83 is –130±36 14C yr, which is consistent (on the basis of a χ2 test with df=37) with the –47±52 14C yr of Russell et al. (Reference Russell, Cook, Ascough and Scott2015). It is worth pointing out here that the T statistic for this grouping is very close to the critical value (T=52.000 and χ2 0.05=52.192, respectively). The value of GUN-177/83 can be used for the later Medieval period in coastal waters of Scotland, corresponding to a later date than the previously available range of ΔR values available for the Holocene period in Scottish waters.
Issues of Taphonomy in Calculation of MRE and ΔR Values Using Archaeological Samples
While the 14C ages of samples from NO-14 are internally consistent within the groups of terrestrial and marine material on the basis of a χ2 test, the two groups of sample ages show a large difference of 2361 14C yr, with the younger samples in this instance (with a weighted mean of 5085±18 14C yr) being the marine shell samples. A negative ΔR value on the order of –1000 14C yr would mean substantially higher 14C content in the oceans than in the atmosphere. While this may be a future prospect in the field of 14C measurement due to the high input of fossil fuels to the atmosphere (Graven Reference Graven2015), it is highly unlikely to have been a feature of past environmental systems on the timescale of the 14C method. It is therefore most likely that the discrepancy in 14C ages at NO-14 is due to issues of taphonomy, namely postdepositional disturbance of a context into which younger (marine) material was entrained. The contexts from which ΔR values were to be determined in this study were carefully selected on the basis of no apparent evidence of such post-depositional mixing; therefore, the 14C ages from NO-14 serve as an example of the need for multiple measurements from contexts, not only for determination of ΔR values, but for contexts where dating is critical to archaeological interpretation, and where material returns an anomalous 14C age contrary to expectations. The experience of NO-14 provides a possible explanation for another of the ΔR values in Russell et al. (Reference Russell, Cook, Ascough and Scott2015) that did not pass the overall χ2 test; Scatness, context 543. In this instance, the calculated MRE was 59±40 14C yr and ΔR=–320±35 14C yr at AD 252–401. Such an extreme negative ΔR may well be explained by intrusion of younger marine material into a context at a later calendar date than when the terrestrial material was deposited. This emphasizes the need for a program of MRE/ΔR assessments for a region in order to obtain a correction value that is accurate as well as precise. The issues of taphonomic bias will always be present on archaeological sites, although these can be mitigated by techniques such as the multiple paired sample approach to MRE/ΔR quantification (cf. Ascough et al. Reference Ascough, Cook and Dugmore2009).
CONCLUSIONS
We present new determinations of the marine 14C reservoir effect (MRE) for key periods in Scottish history and prehistory. We calculate ΔR based on two different methods, the more commonly used intercept method (cf. Russell et al. Reference Russell, Cook, Ascough and Scott2015) and the probability density function method (cf. Reimer and Reimer, forthcoming). The findings were that ΔR values were indistinguishable using the two methods, with a maximum difference of 12 14C yr; however, confidence intervals are slightly larger when using the PDF method, making this the more conservative of the two. We present an interpretation of recalculated values for the earliest period of the Holocene for which MRE values are available. The latter data indicate that in the early Holocene, during the Mesolithic period, MRE/ΔR values were slightly higher than values obtained for the remainder of the Holocene, with a weighted mean ΔR=64±41 14C yr. The new values presented also relate to the latest Mesolithic period in western Scotland, for which no data were previously available. These data suggest a MRE that is slightly higher than values obtained for the remainder of the Holocene, where ΔR=–126±39 14C yr. These values can be used for calibration of samples where the measured marine ages are in the range 5910–5700 14C yr BP, and which are geographically close to the sampled sites. For the later Medieval period, values from the Isle of Lewis indicate a ΔR of –130±36 14C yr for AD 1457–1632. This is consistent with previous ΔR determinations for the period 3500 BC–AD 1450 where a weighted mean ΔR of –47±52 14C yr was determined (Russell et al. Reference Russell, Cook, Ascough and Scott2015). Underlying reasons for the early variations in ΔR that are observed in the data presented here remain elusive, although the 8.2ka BP cold event and associated flux of freshwater to the surface Atlantic Ocean is a possible explanation for the slightly elevated ΔR values. Finally, the findings of this study strongly emphasize the benefit of a program of 14C dating, rather than individual, isolated dates, when seeking accurate chronological information for archaeological deposits, particularly when quantifying the marine 14C reservoir effect, in any geographic area, for any time period.
Further research to build upon the results of this study has the potential to yield valuable insight into the dynamics of MRE and ΔR values in the North Atlantic. Despite the wide temporal range of the values, data are lacking for ΔR in several time periods through the Holocene (e.g. 6000–5000 BC). Sites where a wide variability in MRE/ΔR appears over short timescales (e.g. Carding Mill Bay) warrant more investigation to properly understand this variability. Values from the Iron Age to the Medieval period (i.e. 200 BC–AD 1600) show a variability in ΔR values of between +100 to –200 14C yr. Further research could usefully examine whether this variation is replicated in earlier time periods, in order to improve understanding of the range in values that can be expected for a single geographic area. Finally, future determinations of MRE and ΔR values should focus on periods where large-scale changes in oceanographic changes, particularly salinity, are known to have occurred (e.g. the Little Ice Age and 8.2ka BP cold event), in order to examine whether these changes occur concurrently with specific MRE/ΔR values.
ACKNOWLEDGMENTS
We would like to thank Historic Environment Scotland for providing the range-finder dates from Guinnerso and Tràigh na Beirigh 1 (SUERC-33731, SUERC-33732, OxA-8482, OxA-8483, SUERC-33736, SUERC-33737) and for funding the excavation and post-excavation at Guinnerso. The excavation and post-excavation research at Northton and Tràigh na Beirigh 1 & 2 is funded by Durham University, Historic Environment Scotland and US National Science Foundation (Award Numbers: 0732327 and 1202692). Rosie Bishop is thanked for the detailed identification and recording of the hazelnuts from the Mesolithic sites. We thank Claire Burke (University of Glasgow) and Elaine Dunbar (SUERC) for analytical support in producing radiocarbon ages.
