AI looks for a manuscript’s history in the leaf itself

Turning a palm leaf into a book began well before anyone wrote on it. Depending on local practice, makers boiled and dried the leaves, smoothed their surfaces and treated them with oil. A metal stylus cut the script; dark material rubbed into the grooves made it visible.[2]

Those operations leave a second record. Laura Gallardo and colleagues investigated it with X-rays and machine learning in npj Heritage Science on 6 October.[1]

Millions of palm-leaf manuscripts survive across South and Southeast Asia, according to the Palm-Leaf Manuscript Profiling Initiative. Studying their materials calls for more than one instrument: the programme combines methods including DNA analysis, proteomics, microscopy and spectroscopy with scholarship on texts and manuscript production.[3]

Reading what preparation changed

These leaves carried religious and literary works, technical and legal texts, accounts and correspondence. Their subsequent lives also changed the substrate: humidity, insects, handling and conservation treatments could all affect it. Knowing what a page says cannot, by itself, establish every place it travelled or every treatment it received.[2]

Three X-ray methods reveal different clues: the orientation of fibres, the organisation of cellulose and the elements present. The study compared 85 manuscripts with 75 modern reference materials, using measurements at the DESY and ESRF synchrotrons without taking samples from the objects.[4]

But different measurements from the same object could appear in both training and validation. Accuracy on wholly unfamiliar manuscripts remains unestablished. The strongest result separated historical manuscripts, unprocessed leaves and prepared writing supports. Origin labels can also mix geography with age and preservation.[1]

A biography with several authors

A smaller 2025 study had already explored portable infrared spectroscopy on eleven manuscripts. Its method reduced complex spectra to their main patterns with principal component analysis. It also tested material treated with lemongrass oil, because a conservation treatment can become part of the fingerprint a classifier sees. This is a central historical problem: the surviving object combines the original plant with everything people and time have done to it.[2]

The wider initiative offers a route back to the collections. Its 2024–25 programme planned a transportable container laboratory in Puducherry, working with the French Institute of Pondicherry and the École française d’Extrême-Orient. Those partner collections have been on UNESCO’s Memory of the World register since 2005. Bringing complementary methods into that setting allows material questions to meet the expertise and catalogues surrounding the manuscripts.[3]

The useful ambition is a richer catalogue entry: evidence about the plant, preparation and care alongside script, language and content. A material measurement becomes more valuable when it helps a scholar ask a sharper historical question. The handwriting is one record. The leaf carries another.

Report an error or suggest a correction