The best preserved individuals were selected for isotope analysis of diet (Table 2). These represent 21 individuals including four subadults: two young children aged between 1.5 and 2.5 years old, one child between 7 and 11 years old and one adolescent (12–16 years old), and 17 adults including six females, four males and seven adults whose sex was undetermined.


Methods
Protein food intake may be reconstructed by means of stable carbon and nitrogen isotope ratios in human bone collagen (δ13Ccol and δ15N). Stable carbon isotope ratios allows detection in human diet of the proportion of C4-plants versus C3-plants (Katzenberg et al., 1995) and the proportion of marine versus terrestrial resources (Schoeninger and DeNiro, 1984) whereas stable nitrogen isotope ratios allow a more direct quantitative estimation of the protein contribution (O\‘Connell and Hedges, 1999; Fernandes et al., 2014). Theoretical nitrogen isotope fractionation may vary from 3 to 6‰ between food and consumer (Bocherens and Drucker, 2003; O\‘Connell et al., 2012) while carbon isotope fractionation is only +0–2‰ (Bocherens and Drucker, 2003) between food and consumer. Collagen carbon isotope ratios also include a + 5‰ offset due to fractionation during carbon incorporation into collagen. Stable carbon isotope ratios in human bone hydroxyapatite (or apatite, δ13Cap), on the other hand, reflect the total diet including ci-1033 and lipids as well as protein and thus can be used to assess non-protein sources in the diet, particularly plant consumption (Ambrose and Norr, 1993; Schwarcz, 2000; Kellner and Schoeninger, 2007; Fernandes et al., 2014). Estimates of carbon isotope fractionation between diet and bone apatite range from 9.5 to 13‰ (Howland et al., 2003; Tykot et al., 2009; Warinner and Tuross, 2009; Fernandes et al., 2012); we adopt an offset of 10‰ (Fernandes et al., 2012). Our overall conclusions would not change substantially if a different value in this range were used.
In such theoretical context, merging stable isotope data from both collagen and apatite appears to be theoretically fruitful in the Pacific islands context, where the consumption of low trophic level items like plants (tubers, fruits) and inshore resources (algae, shellfish) characterized human diet (Katzenberg, 2008; Schoeninger, 2010). Carbon stable isotope ratios from collagen are used to identify the consumption of proteins sourced from terrestrial environments versus those sourced from marine environments (Schoeninger and DeNiro, 1984; DeNiro and Epstein, 1978). Collagen nitrogen isotope ratios decipher the consumption of non-reef fish and animal meat from the consumption of horticultural plants and inshore food resources like algae, shellfish, crustacean, and coral reef fish (DeNiro and Epstein, 1981; Schoeninger et al., 1983). Carbon stable isotope ratios from apatite carbonate highlight the consumption of carbohydrate or lipid rich resources such as taro and yam versus C4-plants and inshore resources (Ambrose et al., 1997). In the Pacific Island context, sulfur stable isotope ratios of collagen are of limited interest owing to the sea spray effect (Kinaston et al., 2014).
While carbon and nitrogen stable isotope ratios measured in bone collagen are routinely employed to reconstruct prehistoric human diets of Pacific populations, particularly so because preservation indicators are well documented (DeNiro, 1985; Ambrose, 1990; van Klinken, 1999), the use of apatite carbon isotope ratios has been challenged because there are no unanimously recognized methods for assessing its preservation (Zazzo, 2014). Some researchers have used Fourier-transform infrared spectroscopy (FTIR) and a crystallinity index to assess alterations in crystal structure of the apatite (Wright and Schwarcz, 1996). Others have proposed a comparison of phosphate and carbonate isotope ratios or the amount of CO2 extracted from each sample to identify possible diagenetic alteration (Iacumin et al., 1996, Iacumin et al., 2014). In a study of Maya human bones from Belize, researchers used CO2 gas yield produced during the combustion of apatite samples in the mass spectrometer (Williams et al., 2005), accepting the range of CO2 yields between 0.6 and 1.3% as indicative of unaltered samples (Ambrose, 1993). In the only relevant study of the preservation of apatite carbonate stable isotope ratios in the Pacific area, Ambrose et al. (1997) examined diagenetic effects in bone from the Marianas by the means of assessing correlation between carbonate carbon concentration (%C) and δ13Cap. Specifically, the correlation of these two values would indicate diagenetic apatite carbonate (Ambrose et al., 1997). More recently, Salesse et al. (2014) conducted a detailed analysis of diagenetic effects on isotope signatures (collagen/apatite carbonate) in bones excavated under controlled environmental conditions in a temperate region (Catacomb, Roma, Ist-IIIrd C AD). Using established collagen criteria (%C, %N, C:N), infrared spectroscopy and radiocarbon dating of collagen-carbonate pairs, the authors demonstrated primary macronutrients where the preservation of the organic phase of the bone was poor, samples with high rates of recrystallisation in hydroxyapatite do not show significant isotopic exchanges (Salesse et al., 2014). Interestingly, similar results were also obtained in studies from tropical environments (Zazzo et al., 2014). These studies suggest that, in addition to the routine use of collagen stable isotope ratios for past dietary reconstructions in the Pacific area, stable isotope data from apatite carbonate have the potential to provide useful insights.