Famine was a regular occurrence in post-medieval, pre-Industrial Revolution Europe [1,2]. Where the population was mainly rural, any factor which reduced the quantity of food crops, whether climate, military action, or pestilence, would have a devastating effect on the prices of food and the people who relied on the crops for their calories [3]. In the Great Famine of 1845–46, an attempt was made by Sir Robert Peel to provide relief for the Irish by the importation of maize (‘Indian meal’) from America. This unfamiliar food was unpopular, difficult to process and cook: its yellow colour and effects on the intestines of the starving Irish led to it being renamed ‘Peel’s Brimstone’[4]. The δ¹³C values of potatoes, a C₃ plant and the main food crop in Ireland prior to the Famine, and maize, a C₄ plant which as a group are largely absent from Ireland at this time, are measurably different and this isotopic shift offers the opportunity to investigate dietary change in a population suffering from documented under-nutrition.

The stable isotope ratios of carbon and nitrogen in bone collagen and bulk dentine collagen have been used extensively over the past thirty years to characterise the childhood and adult diets of archaeological individuals [5,6]. Bone collagen, because of its slow turnover rate, represents a long-term average of the adult diet which can, even in mature adults, date back to adolescence [7]. In contrast, primary dentine is mineralised within 3–4 days of secretion and does not remodel, so represents the diet at the time the tooth was growing [7,8]. Methodological improvements have allowed the analysis of much smaller samples of human dentine, e.g. from a maximum of five per tooth by Fuller et al. [8] to sections of collagen weighing 0.5mg [9], which promise greater temporal resolution. While there are variations in the rate at which dentine is secreted as a permanent human tooth grows [10], the method of measuring 1mm horizontal sections used in this study [11] will produce a rolling average of values. At this resolution, and particularly as the method includes the removal of the circumpulpal dentine to avoid contamination with secondary dentine, there should be no significant effect by assigning the same timing to each section (See Fig 1).

The analysis of incremental dentine has allowed researchers to identify short-term changes in the diet of individuals, and the resulting isotope profiles have been used for the first time for the identification of migrants [11], a detailed analysis of the duration of breastfeeding [12], responses to environmental conditions [13] and potential physiological responses to under-nutrition [14]. The δ¹³C and δ¹⁵N values of body tissues reflect not only the diet of an individual, but can also be used as evidence for a period of nutritional stress when dietary intake is insufficient for the individual's energy requirements [15,16]. Neuberger et al. [17] report the segmental analysis of hair for forensic investigation of 15 adults and a child believed to have died as the result of food deprivation. They found a trend for the δ¹⁵N values of the hair segments to rise as body mass index (BMI weight in kilograms divided by the square of height in metres of the individual) reduced. This is consistent with earlier isotopic studies of the δ¹⁵N values of hair from individuals deprived of food [15,18]. Neuberger et al. [17] also noticed a trend amongst adults for δ¹³C values to be in phase with BMI. This observed drop in δ¹³C was thought to be the result of the breakdown of body fat deposits, which are approximately 3‰ lower than other body tissues such as muscle [19] [17]; in order to maintain the levels of glucose and ketones for energy during a prolonged fast, free fatty acids are obtained from the adipose tissue [20] and may also be used for synthesis of new body tissues in the absence of dietary carbohydrate. This observed fall in δ¹³C in nutritionally-stressed adults complements the findings of Mekota et al. [15] who recorded a rise in δ¹³C and a fall in δ¹⁵N in the hair of patients recovering from anorexia nervosa once re-feeding started and those of Lehn et al. [21] in the identification of unknown individuals with a poor quality diet. In contrast, Neuberger et al. [17] found the δ¹³C and δ¹⁵N co-varied in the same direction in hair from a 10 month-old child, suggesting a different metabolic pathway during food deprivation. Cherel et al. [22] discussed the effect of the fasting of King penguins on their δ¹³C blood plasma values, interpreting a rise which they related to lipid content as denoting protein sparing and lipid utilisation during the fasting period. Although energy-partitioning (the relative proportion of energy derived from protein and fat stores) varies between individuals, and is influenced by the pre-starvation percentage of body fat to lean mass, both stores are utilised during prolonged fasting [23].

In the absence of historical documents, identifying famine in a past population presents challenges for the archaeologist. Archaeological contexts may yield evidence for short-term deviations from the usual diet such as inclusion of unusual plants or alterations to butchery practices [24]. Evidence for a catastrophic event in prehistory may hinge upon demography of the cemetery population [25] or archaeological evidence for a single burial event. The osteological analysis of human remains for evidence of malnourishment such as rickets and scurvy and indicators of systemic stress such as enamel hypoplasia, Harris lines and decreased stature all contribute to the health status of individuals [26]. However, Wood et al. [27] pointed out that those who succumb to a rapid death will not survive long enough for the body to produce any bony markers for the causative agent. Those who die from acute diseases associated with famine such as typhus and dysentery will thus appear to have been healthy [28]. Equally, any change in δ¹³C and δ¹⁵N values resulting from a lack of food and leading shortly to death (which can take as little as 2 months [29]) will not be measurable in bone collagen due to long turnover times and averaging [7]. Collagen turnover in individuals who survived periods of famine will also eventually dilute and obscure any isotopic evidence for under-nutrition. This led researchers to conclude it was not possible to identify famine with isotopic analysis of archaeological bone [30,31]. However, starvation is not the most common cause of death [24]: during periods of nutritional stress the population becomes more susceptible to infections and thus it may be difficult to discriminate between an apparent famine and epidemic disease. Morgan [32] concludes that in the absence of historical evidence, secure dating, and isotopic evidence for dietary physiological changes, the combination of all other available archaeological evidence may still not distinguish between catastrophic death from famine and from epidemic disease, or from a combination of both.

As an historical event, the Great Irish Famine of 1845–52 has been extensively researched and interpreted over the last 150 years. Historians and social commentators have used a wealth of contemporary documents to explore the political, social and environmental factors which combined to turn this particular series of crop failures into a devastating disaster. However, Orser [33] carried out the first archaeological excavation of 19^th-century Great-Famine-period dwellings in Ireland in rural County Roscommon and made the point that notwithstanding all the contemporary documents we know very little about the rural poor, and there have been few subsequent opportunities to examine the archaeological record from this period in Ireland. The excavation of the Kilkenny Union workhouse famine cemetery by Margaret Gowen & Co. Ltd. from January to June 2006 allowed careful and respectful analysis of the remains of 970 individuals. The minute books for the workhouse indicate the burial ground was in use between August 1847 and March 1851. The workhouse records from this period, and osteological interpretations of the health and disease of the famine cemetery population, have been published by Geber [34,35] and constitute an insightful analysis of the experiences of the Irish poor at this time. This site represents a unique opportunity to examine the effects of an historical famine on the victims both in terms of their observable bony pathology and of stable isotope analysis of their skeletal tissues to investigate diet and physiology recorded in the δ¹³C and δ¹⁵N values compared with the historical evidence. Such data from people who were nutritionally deprived to such a degree is extremely rare and the opportunity to use their untold stories is of significant import for the assessment of both modern and ancient individuals where nutritional stress may be suspected and soft tissues are not available.

A single permanent tooth was collected from 20 individuals excavated from the Kilkenny Union workhouse famine cemetery, designated numbers KUW 1 to 20. The stable isotope values from incremental dentine collagen were compared with those from the bulk rib bone collagen reported in Beaumont et al. [36]. These individuals were chosen to provide a range of ages and sex, but represent only a small sample of the 970 individuals recorded. Permission was granted for the scientific analysis of these individuals by the Irish Antiquities section of the National Museum of Ireland. All the skeletal remains were re-interred at the Famine Memorial Garden at Goods Shed Square, McDonagh Junction shopping centre, Kilkenny, Ireland, in 2010.

The dentine isotope ratios from the adults reflect their childhood diet which is known from historical sources to have been extremely restricted and dominated by potatoes supplemented with a small amount of animal protein. The dentine δ¹³C values for the adult individuals range between -21.7‰ and -19.5‰ (x = -20.5 +/- 0.3‰ 1sd, n = 13); the δ¹⁵N values range between 9.3‰ and 12.7‰ (x = 11.2 +/- 0.5‰ 1sd, n = 13). The intra-individual range of isotope ratios is smaller: δ¹³C ≤ 1.6‰; δ¹⁵N ≤ 2.2‰. Although small, these ranges exceed analytical uncertainty (i.e. +/-0.4‰ 2sd) and the δ¹³C and δ¹⁵N values co-vary as would be expected during minor changes in dietary protein consumption during childhood and adolescence. For example, KUW 1 exhibits low variation (δ¹³C = 0.8‰; δ¹⁵N = 1‰) in dentine isotope ratio profiles between the ages of 0.5 and 9 years but an incremental δ¹⁵N rise of 1‰ (which exceeds analytical uncertainty at 2sd), between the ages of 5 and 9 years (Fig 2). The progressive nature of the rise suggests this is a real if small change and not merely random analytical error. Whilst the δ¹⁵N dentine (childhood) profile is analytically indistinguishable from the rib bone (adult) value, this is not the case for δ¹³C which shows a ~3‰ increase from c. -20‰ in childhood to -17.2‰ in later life. The rise in δ¹³C is not accompanied by a rise in δ¹⁵N and thus is most likely to be recording a change from a predominantly C₃ potato-based childhood diet to a diet including C₄ maize at some time between the age of 9 years and death more than 13 years later. Additional individual plots for adults from Kilkenny Union workhouse which show that their childhood δ¹³C values reflecting a C₃ potato-based predominantly range from -21‰ and -20‰ are presented in S1, S2 and S3 Figs.

In contrast, the dentine isotope profiles of the seven children who died whilst their teeth were forming, and which thus track diet up to death with increased temporal resolution, display much larger variations in δ¹³C and δ¹⁵N (Fig 3). The δ¹³C values for the children range between -21.4‰ and -14.7‰ (x = -19.5 +/- 0.8‰ 1sd, n = 7); the δ¹⁵N values range between 9.2‰ and 14.3‰ (x = 10.9 +/- 0.5‰ 1sd, n = 7). The intra-individual range of isotope ratios is the same for δ¹³C at ≤ 6.7‰ and marginally smaller for δ¹⁵N at ≤ 4.6‰. The large shifts in the children’s δ¹³C isotope profiles demonstrate that, with the exception of KUW 10 (a nine year old child), they changed their main calorie intake from C₃ potatoes to C₄ maize at some point prior to death. For KUW 4, 11, 13, 14, 16 the δ¹³C profiles were rising immediately prior to death. This is consistent with the documentary evidence relating to the maize-based diet provided in workhouses during the Great Famine [3]. The combined δ¹³C and δ¹⁵N profiles of these children show that, as expected, the maize-based diet increases the δ¹³C values with opposing covariance of the δ¹⁵N values (i.e. the fall in δ¹⁵N occurs at the same time as the rise in δ¹³C), reflecting the change to the C₄ plant-based diet (i.e. the maize pattern) and perhaps suggesting a maize-based diet results in lower δ¹⁵N values. In Fig 3 the profiles from the children are aligned with two known historical events: the introduction of maize imported from North America in March 1846 and the closure of the cemetery in March 1851. Where possible, the profiles are matched to the introduction of maize by aligning the first rise in δ¹³C with March 1846. In two cases (KUW 10 and KUW 12) this meant that the profile continued to death beyond the closure of the cemetery, which was not possible. When those profiles are aligned with death at 1851, the rise in δ¹³C appears to predict the dietary change, showing that there is some error in the calculation of the age assignment of the dentine sections, or there is averaging due to the overlapping developmental structures (see Fig 1). However, in all cases there is a rise in δ¹⁵N just prior to the dietary change to maize (Fig 3). Moreover, the δ¹⁵N values in most children rise just prior to death. The rises in δ¹⁵N are not due to a higher trophic level diet, as δ¹³C should also rise (although to a much smaller extent); in all cases δ¹³C values remain stable or fall, showing the opposite pattern of opposing co-variance (i.e. the famine pattern). Dietary explanations for a rise in δ¹⁵N values without a corresponding rise in δ¹³C values such as the consumption of freshwater fish or meat from animals raised in a forested environment should be considered but these have not been historically documented for this population. Thus, this pattern appears to be evidence for a period of nutritional stress prior to the introduction of maize as a famine relief food. This is consistent with the findings of Neuberger et al., Lehn et al. and Cherel et al. [17,21,22] that the recycling of body fat stores causes a fall in δ¹³C values in body tissues. Huelseman et al. [40], during their controlled feeding experiment where terrestrial foods were replaced with marine, found that the rise in δ¹⁵N values in hair keratin was recorded more rapidly than for δ¹³C. Furthermore, the increase in δ¹⁵N was greater in the individuals with a higher rate of physical activity suggesting that the rise as a result of dietary change was accompanied by a rise associated with nutritional stress.

Both patterns of opposing co-variance are visible in the profile of the juvenile KUW 4 (Fig 4) commencing with low δ¹³C: high δ¹⁵N of the famine pattern between the ages of 4 and 8 years of age and switching to high δ¹³C: low δ¹⁵N of the maize pattern between 8 years of age and death. As with the adult KUW 1 (Fig 2), the δ¹⁵N value for the rib bone correlates well with the dentine values (bone δ¹⁵N = 9.7‰, mean dentine δ¹⁵N = 10‰), and they are indistinguishable in the last c. 5 years of life. The δ¹⁵N value peaks at 10.7‰ at c. 6 years of age which is only 1‰ different from the rib value but is preceded by an incremental rise and followed by an incremental fall suggesting again that this is a subtle but analytically significant change. The shift in δ¹³C between dentine and rib is large: the dentine has a mean value of -21‰ (n = 10) between the ages of 2 and 9 years but the δ¹³C rib value of -16.2‰ clearly reflects the shift to a C₄ maize diet between the ages of 8 and 9 years. Whilst the rib value falls short of the peri-mortem dentine peak of -14.7‰, showing it was still in the process of equilibrating with the C₄ diet at death, it sits between the means of the last four increments and the last three increments, i.e. -16.7‰ and -15.9‰ respectively, suggesting the rib value of this c. 13 year old child averages the last c. 3 years of dietary inputs. Additional individual profiles for juveniles from Kilkenny Union workhouse presented in S3 Fig show that, where there has been a C₃ to C₄ dietary change prior to death, rib and final dentine increments can record different δ¹³C values where, as in the case of KUW 4, the rib appears to be lagging behind the dentine (e.g. KUW 11,13,16) or, somewhat inexplicably, predicting future dentine values (e.g. KUW 14).

The difference between the δ¹³C value of individual adult rib and mean dentine (Δ¹³C) varies from 0.2‰ to 2.8‰ but is, in all cases, positive, i.e. later life values are higher than childhood values (Fig 5) although there is no significant correlation with age at death and Δ¹³C (Pearson r = 0.43, p-value = 0.15). The δ¹³C mean of the ribs is significantly greater than the mean of the bulk dentine (paired t-test p < 0.05, n = 13) and the data provide strong evidence that short-term consumption of C₄ maize has been variably recorded in the rib collagen through bone turnover during the last few years of life whilst the dentine retains the C₃ potato-based isotope values of the childhood diet (Fig 5 and S3 Fig). The retention of early life C₃ isotope values in the adult dentine is supported by the fact that virtually none of the adult dentine δ¹³C values rise above -20‰, the dentine values for adults are significantly less variable than those of the children (adult dentine mean -20.4‰ ± 0.4‰ 1sd; children dentine mean 19.6‰ ± 1.5‰ 1sd, p < 0.05) and that the children's dentine mean is significantly greater than the adult dentine mean (t-test p < 0.05).

As indicated above, the consumption of maize by the adults in the Kilkenny Union workhouse is recorded variably in the ribs and whilst there is no statistically significant correlation with age at death in this relatively small group, it is difficult to conclude what other factors such as onset of maize consumption, activity, nutrition and health may have impacted on the visibility of maize consumption in the adult bones. Equally, the transition from potatoes to maize visible within the dentine δ¹³C values of the children also appears to take several years (e.g. Fig 4) because each sample of dentine, although of far higher temporal resolution than bone, still does not represent a discrete time period but rather a running average due to current sample size requirements and the pattern of appositional dentine growth (Fig 1) [11]. As a result, a step-change in diet from C₃ to C₄ will manifest as a rising profile starting with the increment that includes the first C₄ influenced dentine in the innermost layer. This dentine will also be forming simultaneously in outer dentine layers nearer the tooth apex. If these layers continue to build up with only C₄-influenced dentine the profile further down the tooth, if not arrested by death, will plateau at the new higher δ¹³C value. However appositional growth means that the amplitude of a temporary dietary change within a profile will be attenuated and the peak flattened as is suggested for the δ¹⁵N peak of KUW 4 that occurs between the ages of 5 and 7 years (Fig 4). Here, it is highly likely that the visible peak does not represent the highest short-term δ¹⁵N dentine value because after this dentine layer was laid down, additional layers would have been added with lower δ¹⁵N values reducing the averaged value. The true height of the peak is therefore no longer visible at the current sampling resolution as KUW 4 survived this period of stress. This strongly implies that a period of famine may manifest as what might be reasonably considered a fairly minor (i.e. ~1‰) rise in δ¹⁵N. A further interpretive complication is that increments at the beginning and the end of a tooth profile may record more extreme values because they are subject to only one-way averaging, i.e. dentine formed either solely after or before the increment in question not both, and thus they will have a higher temporal resolution. In the case of the final increments of incomplete tooth roots, such as that of KUW 4 and the other juveniles in this study, the temporal resolution will be even greater because the dentine has not yet reached its final thickness. For most incomplete teeth, the final increment is extremely thin consisting of outer dentine layers only, and thus its isotope ratios will not be ameliorated by dentine that would have been laid down in the future if the individual had survived and the tooth continued to completion. Turning this around, it also implies that the onset of a dietary change from C₃ to C₄ will be heralded at an earlier age in the profiles presented here than it actually occurred because the inner, later forming layers of dentine will cause the profiles to rise. These observations suggest that similar dietary or physiological changes may manifest differently in dentine isotope profiles depending on whether the change occurred within a completed tooth root and was survived, or at the end of a profile in a tooth where development was arrested at death. Temporary changes in δ¹³C as a result of a change in diet or increases in δ¹⁵N as a result of famine may, therefore, be far more subtle and of suppressed amplitude than might be anticipated or indeed obtained from the incomplete tooth roots of children. Thinning the dentine of fully formed teeth from the pulp cavity prior to sectioning should reduce the impact of later appositional deposition of dentine and thus increase the temporal resolution of the isotope profiles.

The Great Irish Famine of 1845–52 was not the first major famine in Ireland nor was the use of maize as a relief food in March 1846 the first major importation of maize to Ireland. Crawford [41] demonstrates that maize was imported in quantity in 1801 and again in 1827 to supply cheap food to the Irish poor. During 1827, when there were widespread crop failures, Humphrey O’Sullivan, a schoolteacher in Callan, County Kilkenny, records in his diary that 'Indian meal' imported from America was being distributed in an attempt to “keep down the cost of living for the poor” [41]. Whilst it is only possible to speculate about the actual cause of death for the people buried in the Kilkenny Union workhouse cemetery because many individuals died as a result of an inability to fight infection [42], the adult profiles show that some of this workhouse population, who represent the poorest members of society at this time, appear to have suffered from earlier periods of famine which have been recorded in dentine formed during their childhood. The dietary history of Ireland since 1500 as documented by Clarkson and Crawford [3] included many periods of food shortage whether arising from conflicts, extreme weather conditions, or political mismanagement. It was always those with the least entitlement to food who suffered the most [43]. It is possible, therefore, that older individuals in the Kilkenny cemetery had survived earlier periods of food shortages in the half century preceding their death in the Kilkenny Union workhouse. KUW 5, an adult female whose teeth would have been forming during the 1827 famine, shows both patterns of opposing co-variance in her dentine profile (Fig 6): the maize pattern between the ages of 4 and 7 years, when the δ¹³C values temporarily rise above -20‰ coupled with a 1.4‰ fall in δ¹⁵N, followed by the famine pattern where a 1.1‰ fall in δ¹³C values to within the C₃ range is coupled with a 1.6‰ rise in δ¹⁵N denoting a further nutritional stress event between the ages of 8.5 and 12 years. The later-life values of the rib suggest both δ¹³C and δ¹⁵N are raised above those expected for a nutritionally sufficient potato-dominated C₃ diet in the years before death.

This study is the first to use the stable isotope analysis of nitrogen and carbon in human dentine to identify changes related to a well-documented historical period of famine. These data are a poignant reminder of the extreme conditions suffered by the children buried at Kilkenny, but also clearly demonstrate that the intervention provided by the introduction of maize allowed many to recover: their nitrogen isotope ratios return to dietary rather than stress levels, and they survived beyond the dietary change. The ability of the human body to adapt to nutritional deprivation and survive, returning to homeostasis, is clear within these profiles.

Although the change in diet from a C₃ to a C₄-based diet is recorded in the isotope ratios of the adult rib bone collagen, it is difficult to resolve the age at which this occurred because of averaging as a result of bone turnover. Using evidence from the dentine collagen δ¹³C and δ¹⁵N profiles and the historical records for both the importation of maize and the timing of cemetery closure we have been able to identify more precisely when in their lifetime the children from this study who died during the Famine period received relief food. The raised δ¹⁵N and lower δ¹³C (famine pattern of opposing covariance) in the dentine profiles of the children is marked by the return to lower δ¹⁵N values and higher δ¹³C in the maize pattern of opposing covariance. Short-term δ¹³C and δ¹⁵N changes in dentine resulting from physiological, in this case specifically nutritional, stress has not previously been demonstrated in an historical population where the records of the period of famine can be related to individuals who suffered. Further work needs to be undertaken to refine the ageing of each increment, both chronological age and the period of time represented, and whether partially mineralised teeth need ageing differently to complete ones. Nonetheless, this study shows that using incremental dentine collagen δ¹³C and δ¹⁵N analysis it is possible to identify periods of physiological stress such as famine in both adult and juvenile skeletons if it occurred during tooth development, i.e. up to approximately 23 years of age. This could have important forensic and archaeological applications for the identification of populations and individuals for whom nutritional stress may have contributed to their death.