Somewhere in southwestern Sri Lanka, roughly 20,000 years ago, a person chewed through whatever they found in the rainforest that day. Fruits, probably. Possibly breadfruit or candlenut. Certainly some animal protein. The exact menu is lost. What wasn’t lost, it turns out, was the chemical record of that diet, preserved in the crystalline enamel of their teeth.
A new study in Nature Ecology & Evolution1 by Bourgon and colleagues has analyzed zinc isotope ratios from the enamel of 24 human specimens and 57 faunal samples from three well-documented Sri Lankan cave sites: Fa-Hien Lena, Batadomba-lena, and Balangoda Kuragala. The combined record spans roughly 20,000 to 3,000 years ago, taking in the tail end of the Last Glacial Maximum, the Pleistocene-Holocene transition, and the period immediately before domesticated crops appear in the island’s lowland zones. What that record shows is a slow, measurable drift in the human diet toward plants, long before anyone was farming.
The method deserves some explanation, because it’s doing something that wasn’t possible until recently. Standard dietary reconstruction in archaeology relies on stable nitrogen isotopes in bone collagen, which track trophic level reliably. The problem is that collagen degrades. In tropical environments, it’s often gone entirely. Sri Lanka’s cave sites preserve some of the earliest Homo sapiens fossils found anywhere in South Asia, but those fossils are old enough, and wet enough in their burial context, that collagen is frequently absent or unreliable.
Zinc isotopes in enamel are different. Enamel is dense and highly mineralized, and it preserves the biogenic zinc signature even under conditions that destroy organic tissue. The key property is directional: herbivores show the highest δ⁶⁶Zn values, carnivores the lowest, and omnivores fall in between. As trophic level increases, δ⁶⁶Zn decreases. The mechanism involves fractionation across tissues and the fact that animal foods contain more zinc and promote greater absorption than plant foods. This last point matters for interpretation. Because animal protein has an outsized effect on zinc isotope ratios relative to its caloric contribution, even moderate meat intake can pull values toward the carnivore end of the scale. Which means that when humans sit consistently in the intermediate range, the contribution from plants must have been substantial.
The faunal dataset the team assembled provides the interpretive scaffold. Deer, sambar, barking deer, rhinoceros, elephant, wild boar, flying squirrel, porcupine, bat, civet, and various primate species are all represented. This spread gives a local trophic structure against which the human values can be plotted. The human specimens consistently sit between the herbivores and the macaques.
The macaques are interesting. Macaca sinica, the toque macaque endemic to Sri Lanka, shows very low δ⁶⁶Zn values, overlapping with what you’d expect from obligate carnivores, despite being an omnivore that eats a substantial proportion of plant food. The low values probably reflect its habit of consuming eggs, insect larvae, small reptiles, birds, and mammals alongside fruit and leaves. Because obligate carnivores are absent from Sri Lanka’s cave fauna, M. sinica ends up occupying the low end of the isotopic range and functions as a carnivore analogue in the system. The team validated this by comparing the spacing between macaque values and herbivore values against equivalent data from mainland Southeast Asian sites where carnivores are present, and the match is close enough to support the interpretive framework.
Against that baseline, humans sit in the middle. That alone is not surprising, given what the archaeological record at these sites already suggests: microliths, bone tools, and what is currently the earliest confirmed bow-and-arrow technology outside Africa all point to skilled, organized hunting of arboreal and semi-arboreal mammals. Mollusc shells, breadfruit, and candlenut show up too. Mixed subsistence seems obvious in hindsight.
The direction of change over time is less obvious. That’s the study’s more striking finding.
