New perspectives on nutritional stress—first carbon and nitrogen stable isotopes of pig remains from historic starvation experiments

This exploratory study presents the first isotopic data obtained from skeletal remains of pigs that are known to have suffered severe nutritional stress, compared to a control group of pigs from the same litters. A pilot investigation of pig remains from 20th century feeding experiments is conducted. The small dataset (n = 10) shows that potential isotopic signals of starvation are obscured by the different dietary compositions of both test groups. This makes it unlikely that bulk stable isotope analysis of archaeological faunal remains can be used to identify nutritional stress in a setting of unknown diet and starvation status.. Fattened individuals had, compared to their starved siblings, δ 1 ⁵N values ranging from −0.73‰ to +0.16‰ (mean: −0.36‰), and δ 13 C values ranging from −0.24‰ to +0.03‰ (mean: −0.12‰). Differences between two pig breeds are observed, indicating that breed could be a currently overlooked factor for interpreting variation in archaeological intra-species isotopic values.

Stable isotope analysis of bone and dentine collagen is widely used to reconstruct and interpret life histories of humans and other animals in archaeological studies (e.g., Britton, 2017; Sharpe and Krigbaum, 2022). Although various applications are well established, other approaches need more investigation. Specifically, using stable isotope analysis for identifying nutritional stress in bioarcheological remains is not yet possible with confidence.

Previous studies have observed that prolonged and severe nutritional stress leads to higher δ 15 N values in body tissues of humans and animals, which is attributed to the catabolism of body proteins to compensate for a lack of nutrients (Beaumont and Montgomery, 2016; Doherty et al., 2022; Eerkens et al., 2017; Mekota et al., 2006; Neuberger et al., 2013). and an increased excretion of 14 N (Doi et al., 2007; Lee et al., 2012). An inverse correlation was sometimes observed in δ 13 C values, which is thought to be caused by the utilization of 13 C-depleted fat deposits (Beaumont and Montgomery, 2016; Doherty et al., 2022; Doi et al., 2017; DeNiro and Epstein, 1977; Kiljunen et al., 2006; Neuberger et al., 2013). Generally, these patterns can vary between organisms (MacAvoy et al., 2006) and tissues, whereby tissues with higher turnover tend to show stronger effects (Cherel et al., 2005; Doherty et al., 2022). A meta-study of fasting experiments proposes that starvation can lead to variation between −0.82‰ to 4.30‰ for δ 15 N and −1.92‰ to 2.62‰ for δ 13 C values, whereby especially the variation in δ 13 C was difficult to predict and is thought to be not well understood (Doi et al., 2017).

The nature of these isotopic shifts caused by starvation were investigated with a range of controlled studies. Some ecological approaches focused on naturally recurring periods of starvation, for example in molting penguins (Cherel et al., 2005) or hibernating ground squirrels (Lee et al., 2012). Controlled studies have been conducted on birds (Hobson et al., 1993; Kempster et al., 2007), fish (Gaye-Siessegger et al., 2004), scallops (Régnier-Brisson et al., 2024), larvae (Doi et al. 2007; Tibbets et al., 2008) and rats (Robertson et al., 2017). Animal welfare concerns have sometimes caused these studies to be terminated before clear isotopic patterns became visible, as these are assumed to require severe and prolonged periods of nutritional stress (Doi et al., 2017; Kempster et al., 2007). In those controlled studies that have found a significant impact of nutritional stress on isotopic ratios, they vary strongly between organisms and tissues, from +0.24‰ to +1.68‰ for δ 15 N and −0.05‰ to +0.3‰ for δ 13 C values (Cherel et al., 2005; Doi et al., 2007; Hobson et al., 1993; Régnier-Brisson et al., 2024; Robertson et al., 2017).

Other studies have utilized the isotopic changes attributed to nutritional stress through archaeological investigation of famine burials (Beaumont et al., 2013; Beaumont and Montgomery, 2016; Drtikolová Kaupová et al., 2026; Walter et al., 2020), or skeletal remains with paleopathological indicators of malnutrition (DeWitte et al., 2025; Doherty et al., 2023; Godde et al., 2025; Riccomi et al., 2024). A limiting factor in these studies was the less controlled environment compared to the experimental approaches. As many different factors could influence changes in stable isotope ratios, some of these studies found either indecisive (Riccomi et al., 2024), or even the opposite of the expected patterns. Specifically and contrary to expectations, remains from famine burials were found to have −0.2‰ to −0.6‰ lower δ 15 N values compared to other burial types. This was interpreted as being indicative of social status, whereby higher status allows for a protein rich diet and is generally protective against famine events (Drtikolová Kaupová et al., 2026; Godde et al., 2025; Walter et al., 2020).

These previous lines of evidence to establish the isotopic changes caused by nutritional stress are not fully suitable to understand periods of deprivation in past farming systems. The reason is the vast differences with which nutritional stress is expressed isotopically between species and tissues. Therefore, zooarchaeological studies aiming to identify nutritional stress in archaeological remains of livestock currently have no suitable baseline to refer to. Even studies with other mammals cannot serve as comparison because the variation of the isotopic integration of diets between mammalian species is too different (MacAvoy et al., 2006), leading to the huge variability of isotopic shifts caused by starvation observed in previous studies.

What is currently missing is data from controlled studies, conducted on the same taxa and skeletal element combination that zooarchaeological studies want to investigate, as only such data can form a suitable baseline. As such, this paper contributes to the field of stable isotope studies in zooarchaeology by utilizing pig remains from early 20th century feeding experiments, which are more closely resembling past farming systems than any material previously analyzed to investigate the isotopic changes caused by nutritional stress. This study aims to (1) investigate whether nutritionally stressed pigs show higher δ 1 ⁵N values relative to fattened siblings; (2) test whether δ 13 C values follow the expected pattern of opposing covariance; and (3) assess if the impact that nutritional stress has on isotopic values is even visible if factors like diet composition, breed and sex vary between the starved and fattened groups. To achieve these aims, the conditions of the experiments and their results regarding the body mass development of the pigs are presented together with a pilot dataset of carbon and nitrogen stable isotope values deriving from bone and dentine collagen of six individuals.

Selected material and background

In 1910, Simon von Nathusius, director of the Livestock Garden at the Martin-Luther University in Halle (Germany), began conducting starvation and fattening experiments on pig siblings to investigate the influence of diet on physiological development. He obtained two pregnant pure-bred sows, one Hannover-Braunschweiger (a robust heritage breed) and one Berkshire (a highly specialized breed). Three weeks after weaning, their litters were divided into test groups. Of the Hannover-Braunschweiger litter, two animals were assigned to a fattening and three to a starvation group, while the Berkshire litter was split into two groups of four animals each. Food rations were calculated based on total living weight for each group, usually adjusted weekly, and provided to the group as a whole rather than to individual animals. Therefore, members of a group competed for the available food, providing a good analogue to livestock farming practices in the past. The main difference in their diets was that the starvation group received only one quarter of the peanut-flour and no barley or milk (the exact composition of each group's diet can be seen in Supplementary Table S5). All food was sourced from the same farmers for all test groups; the animals received the same water supply (ad libitum) and were kept in comparable pens. The experiment was terminated after three months and one animal from each test group slaughtered immediately, while the remaining individuals were placed on a normal diet (a slight reduction from the fattening diet) and lived for several more months (Henseler, 1913, 1914; Von Nathusius, 1912) (Figure 1). All carcasses were manually defleshed by the institutes’ taxidermist, slightly heated in a mild soap solution and stored in paper bags in a barn, avoiding frost and humidity over 60%. Likely, no chemicals were used for conservation although this is not documented (ZNS, 2026; Schafberg, pers. comm.).

Six mandibular molars (M 1 s) and four rib bones of six individuals were sampled, encompassing the four individuals which were slaughtered immediately and two more Hannover-Braunschweiger pigs. Of the latter, one M 1, of the former, one M 1 and one rib bone per individual were sampled. See Table 1 for the animals' age-at-death, sex, and weight. The body mass development of the six individuals is illustrated in Figures 2, 3:

The development of the body masses illustrates the severity of the nutritional stress experienced by these pigs. It is noteworthy that the starved pig S hv-br 6 was able to recover to such an extent that it reached c.95% of the body mass of its fattened sibling S hv-br 5 after sharing a diet for some months (compared to c.39% at the end of the experiments).

Methods

The selected teeth and rib bones were mechanically abraded with a handheld Dremel drill, before removing 200–300 mg dentine (M 1 s) or bone (ribs) per specimen. For rib bones, square pieces were sampled from the middle of the bones. For M 1 s, distal ends of tooth roots were sampled, representing the late forming part of the tooth (Tonge and McCance, 1973). Samples were defatted thrice with a 2:1 dichloromethane/methanol solution. No stable isotope ratios could be established for the diet as no diet samples from the experiments remain. Collagen was extracted and purified following the demineralization, gelatinization, ultrafiltration and lyophilization protocols outlined in (Chidimuro et al., 2023). All collagen samples (0.70–0.90 mg) were weighed out in duplicates into 4 × 3.2 mm tin-capsules and analyzed in a SerCon “ Callisto CF-IRMS ” system at the University of Exeter's Environment and Sustainability Institute. Sample data were reported in standard delta per mil notation (δ‰) relative to Vienna Pee Dee Belemnite (V-PDB) (δ 13 C) and atmospheric air (AIR) (δ 15 N) international standards. Duplicate sample pairs were reported as mean values. Two laboratory reference materials [Sigma alanine (EP Reference Standard) and bovine liver] were interspersed within the measurement run to monitor and correct for non-linearity, instrument drift and analytical precision. Differences between groups were tested for significance with a two-tailed Exact Wilcoxon-Mann–Whitney test.

Collagen quality indicators and measurement precision

Collagen yields (dry mass) of 21%–24% and 10%–19% were obtained from bone and dentine, respectively, well above the 2%–4% threshold used to identify problematic samples (Van Klinken, 1999). Dentine %C ranged from 33.91% to 47.62%, and %N from 11.82% to 16.44%. Bone %C ranged from 42.64% to 47.06%, and %N from 14.59% to 15.89%. Atomic C:N ratios in dentine ranged from 3.26 to 3.46 (mean: 3.36), and in bone from 3.37 to 3.57 (mean: 3.46), which is consistent with bone collagen that has not undergone diagenetic alterations (DeNiro, 1985). The measurement precision (2xSD) on both calibration standards was ±0.47‰ or better for δ 15 N and ±0.19‰ or better for δ 13 C. The average difference between duplicate sample pairs was 0.09‰ for δ 15 N and 0.08‰ for δ 13 C. The unabridged values are reported in Supplementary Table S2.

Stable isotope values

The M 1 tooth root formation had not terminated in the four individuals that were slaughtered directly after the experiment, as all roots were still open with thin walls. Because dentine does not remodel once formed (DeWitte et al., 2025; Hillson, 2023), it can be assumed that the roots were formed roughly during the second half of the experiments, considering that a regularly raised Large White pig completes M 1 root formation at around 240 days of age (Tonge and McCance, 1973). In contrast, ribs were chosen to represent a later stage in the individual's lives, which is common in archaeological studies (cf. e.g., Reitsema and Vercellotti, 2012; Walter et al., 2020). However, it is known from human bones that turnover rates can vary significantly within skeletal elements and ribs equally represent a temporal average (Hall et al., 2026; Quinn, 2024). To simplify this issue for the context of this study, it is assumed that the bone collagen represents a later point within the experiments than the dentine collagen, allowing to determine temporal variation in individual life histories.

The pattern visible in the four animals slaughtered directly after the experiments terminated is largely consistent (Figure 4). Among these, values obtained from starved siblings were always higher in δ 15 N compared to their fattened counterparts. This is true for both breeds and values deriving from bone and dentine collagen. Fattened individuals had, compared to their starved siblings, δ 15 N values ranging from −0.73‰ to +0.16‰ (mean: −0.36‰). Of these, only the difference between S bks 11 & 12 (bone) and S hv-br 4 & 3 (dentine) are higher than the doubled standard deviation. Differences in δ 13 C are less pronounced although equally consistent with all samples from starved individuals expressing higher δ 13 C values than their fattened siblings among the animals slaughtered directly after the experiments. Fattened individuals had, compared to their starved siblings, δ 13 C values ranging from −0.24‰ to +0.03‰ (mean: −0.12‰), whereby no difference between any pair was higher than the doubled standard deviation. No statistically significant differences were found when comparing δ 15 N or δ 13 C values from starved to fattened individuals.

The difference between bone and dentine collagen values is not consistent. In three of four individuals, the values from dentine are higher in both δ 13 C and δ 15 N compared to bone-derived collagen from the same individual. However, in fattened individual S hv-br 3, only δ 13 C is higher while δ 15 N is slightly lower. Bone-derived collagen had, compared to dentine-derived collagen, δ 15 N values ranging from −0.44‰ to +0.02‰ (mean: −0.35‰), and δ 13 C values ranging from −0.28‰ to −0.36‰ (mean: −0.33‰).

Generally, the Hannover-Braunschweiger pigs have both higher δ 15 N and higher δ 13 C values than the Berkshires, when comparing values from the same diet and tissue combination between both breeds. Berkshires had, compared to Hannover-Braunschweiger pigs, δ 15 N values ranging from −0.44‰ to +0.04‰ (mean: −0.34‰), and δ 13 C values ranging from −0.48‰ to −0.26‰ (mean: −0.36‰).

The dentine-derived values are very similar for the two siblings that were allowed to live after the experiments. Their δ 15 N values are comparable with those from their sibling S hv-br 4, which was slaughtered at the height of its starvation. At the same time, their δ 13 C values are higher by almost +1.5‰ compared to the dentine values of their earlier slaughtered siblings. Both indicate that these values represent the diet after the experiment was concluded, rendering them incomparable to the rest of the dataset.

The assumption that nutritional stress generally leads to higher δ 1 5 N values is confirmed for both pig breeds, although only in individuals slaughtered during ongoing starvation. The values are only marginally higher, however. It must be considered that one of the main differences between the test groups in these experiments was the difference in dietary composition (cf. Supplementary Table S5). Fattened animals received a much higher amount of peanut flour. Being a leguminous plant, peanuts have lower δ 15 N values than non-leguminous plants by around 3‰ (Chandler et al., 2025). At the same time, the additional milk for fattened animals would lead to higher δ 15 N values. And even if sharing the same diet, δ 1 5 N values in pig rib collagen can vary (Webb et al., 2016). Furthermore, different tissues reflect the diet's isotopic composition to different degrees (Tieszen and Fagre, 1993). Therefore, the observed difference in δ 1 5 N values between the starvation and fattening groups is not based on the effect of starvation alone but likely caused by a mixture of dietary composition and the expected catabolism of body proteins in the starvation group.

No distinct isotopic pattern remains visible in pigs that resumed a shared diet after the experiment, even though a significant portion of the sampled M 1 dentine formed during the starvation period. As it is generally assumed that dentine does not remodel after formation (DeWitte et al., 2025; Hillson, 2023), this is surprising and recommends future investigations with larger samples sizes to explain why no trace of the starvation period remains visible in the δ 1 5 N values of S hv-br 6 compared to its sibling S hv-br 5.

The increase of δ 15 N values is not consistent between tissues. The Hannover-Braunschweiger siblings show a smaller difference in rib collagen but a larger difference in dentine collagen compared to the Berkshires. The generally higher δ 15 N (with one exception) and δ 13 C values in dentine compared to rib collagen may reflect differences in the timing of tissue formation. However, previous research indicates that tissues differ isotopically due to their distinct formation and turnover processes (Doherty et al., 2022; Hall et al., 2026; O'Donoghue et al., 2021), cautioning against purely chronological interpretation. If, however, rib collagen is interpreted as reflecting a later stage of the experiment, this pattern may indicate that Hannover-Braunschweiger pigs adapted better to a low-food diet than Berkshires, which aligns with the observation that starving Hannover-Braunschweiger pigs expressed livelier behavior than Berkshires during the experiment (Von Nathusius, 1912).

Interestingly, Hannover-Braunschweiger pigs consistently show higher δ 15 N and δ 13 C values across all datapoints compared to their Berkshire counterparts. Given the controlled test environment, this might either reflect their different sexes (cf. Table 1; the influence of sex on carbon and nitrogen stable isotope values is e.g., known from rats: Kurle et al., 2014) but could also relate to physiological and metabolic differences between breeds, potentially indicating that breed is a hitherto under-considered factor for interpreting intra-species isotopic variation in zooarchaeological studies.

Contrary to expectations, δ 13 C values were normally higher in starved pigs, although the difference was not significant. This is likely caused by the different dietary compositions, indicating that even if only C3 plants were consumed, the potential impact of starvation is easily obscured by different diets. Together with the δ 1 5 N values, this implies that, in an archaeological setting with unknown diet and starvation status, bulk stable isotope analysis is unlikely to succeed in unambiguously identifying nutritional stress: the impact of different dietary compositions is so strong that even the severe starvation suffered by the analyzed pigs leads to only marginally different isotopic values.