Phytolith extraction followed the method described by Parr et al. (2001) and D’Agostini et al. (2022) with some modifications that considered the specificity of manuscript samples. Every sample was air dried and ashed at 550°C for 3 h in a muffle furnace. After a 12-h cooling period ash was transferred into test tubes to undergo treatment with 10 mL of 10% hydrochloric acid (HCl) for 30 min. One tablet of Lycopodium clavatum (number of spores 20,848 ± 1546; Stockmarr, 1971) was added to each sample at the first step of chemical treatment to enable estimation of the phytolith concentration (amount of phytoliths per one ml of studied material) and content (amount of phytoliths per gram of studied material). After centrifugation at 6000 rpm for 5 min, samples were washed twice with distilled water following additional centrifugation after each washing step. Next, samples were treated with 10 mL of 10% hydrogen peroxide (H₂O₂) for 2-5 h followed by another 5 min of centrifugation. Thereafter, peroxide was decanted and samples were twice washed with distilled water and centrifuged again. In order to possibly avoid additional erosion and dissolution effect, no vortexing was applied and no strong acids were used. We did not wash nor perform any other sort of cleaning on any of our samples before ashing and wet laboratory processing in order to study the surface environmental contamination in all available samples. Fresh, dry, herbarized and manuscript samples were processed separately, on different days in order to avoid possible cross-contamination.

Residues were kept in the fridge (5°C) in distillate water; slides for light microscopy were prepared with sterile liquid glycerin since that allows to rotate the counted micro-objects and ensures better investigation of the phytolith morphology. Permanent slides of the research material were prepared with glycerin gelatin. Microslides were examined under light microscope at a magnification of x400, x600 and x1000 times. In order to separate phytolith with ambiguous appearance and random mineral particles, polarized light microscopy was applied.

A minimum of 300 phytoliths were counted per sample. All phytoliths greater than 2 µm were photographed, described morphologically and morphometrically, if it was needed for diagnostic purpose. In order to estimate levels of old palm material deterioration (i.e., in herbarized and PLM material), degraded and eroded phytoliths as well as silica sand was counted. All amorphous, rectangular and hexagonal silica fragments of unknown nature as well as phytoliths less than 2 µm in each linear dimension were counted together as silica sand. Sand counting was performed in one observation field at the magnification of x400 and then multiplied by the number of observation fields used for the same phytolith sample. Ashing and all wet laboratory preparations were partly performed at the Department of Palynology and Climate Dynamics, University of Göttingen and partly at the Institute of Plant Sciences and Microbiology (IPM) of the University of Hamburg, Germany.

All phytolith types described for fresh, dry, herbarized and manuscript material of both investigated palm species, were divided into seven functional groups according to phytolith morphology and morphometry as well as to the most probable source plant group(s) that was determinate, namely (1) Arecaceae, that comprises phytoliths mostly originating from Borassus or Corypha palm-leaf material, but includes also isolated (i.e., never aggregated) silica bodies of other palms, observed in SEM either strictly on the surface of PLMs or seen as a contamination on the surface of unprocessed leaves. (2) Arecaceae/Zingiberaceae and (3) Arecaceae/Zingiberaceae/Bromeliaceae, seen rarely and randomly on the surface of the unprocessed leaves and often on the PLMs. (4) Musaceae, and (5) Poaceae, mainly registered in the manuscript samples; (6) phytoliths diagnostic for other plants, and (7) non-diagnostic phytoliths. The phytolith morphology and terminology employed here is based on the International Code for Phytolith Nomenclature (ICPN; Neumann et al., 2019), if not stated otherwise.

Phytoliths were identified following morpho-taxonomical guidance of Piperno (2006), Piperno, 2009); Piperno and Pearsall (1998); Pearsall (2011); Chen and Smith (2013), ICPN 2.0 (2019), Piperno and McMichael (2020), and the morphometric studies of Ollendorf (1992); Albert et al. (2009); Fenwick et al. (2011), and Golokhvast et al. (2018). Arecaceae phytoliths were identified following Tomlinson (1961); Sangster and Hodson (1992); Tomlinson et al. (2011); Romain and De Franceschi (2013); Morcote-Ríos et al. (2016); Collura and Neumann (2017); Witteveen et al. (2022); Liu et al. (2023) and literature cited within. In doubtful cases, in order to separate phytoliths from other phytolith types with some similar shape and surface ornamentation, descriptions and illustrations for phytoliths of Orchidaceae (Piperno, 2006; Chen and Smith, 2013; Sharma et al., 2018), Bromeliaceae (Tomlinson, 1969; Piperno, 2006), Cannaceae, Marantaceae, Strelitziaceae, and Zingiberaceae (Kealhofer and Piperno, 1998; Chen and Smith, 2013; Benvenuto et al., 2015; Wang et al., 2022; Dai et al., 2023) and own reference collection material were used.

Musaceae phytoliths were identified as proposed by Mbida Mindzie et al. (2001); Ball et al. (2006); Neumann and Hildebrand (2009) and Chen and Smith (2013). Cannabis sp. phytolith assemblage included shapeless, oval, segmented ovals, and club-shaped phytoliths as well as spikes, all ca. 20-50 µm in diameter and ca. 20-70 µm in length as described by Golokhvast et al. (2018); Table 2 ). The grass silica short-cell phytoliths (GSSCP) diagnostic for Poaceae family were classified as described in ICPN 2.0 (2019) and by Dai et al. (2023). When preservation allowed, genera of Poaceae were recognized (e.g., as in Zhao et al., 1998; Yost and Blinnikov, 2011; Gu et al., 2013, Gu et al., 2016; Huan et al., 2015; Tao et al., 2019; Wang et al., 2022; Cordova, 2023). If the phytolith morphology suggest a certain identification but does not look exactly like the reference plant material, phytolith morphotypes are marked with “cf.”

If more than one possible assignment of the phytolith morphological type (morphotype) to the source plant(s) was possible, all plant groups were indicated, e.g., Arecaceae/Zingiberaceae/Bromeliaceae, Poaceae/Cyperaceae. Phytoliths for which more than three probable source plant groups were suggested after detailed morpho-comparative analysis and phytoliths occurring within a large number of plants were assigned to the ‘non-diagnostic phytoliths’. In case of doubt, when for any reason it was impossible to verify morphological assignment (only one phytolith of a certain type, poor preservation, ambiguous appearance) or if it was overall difficult to reveal a source plant, a conservative approach was applied and phytoliths with ambiguous morphology were grouped together with non-diagnostic phytoliths. If the preservation conditions and/or visual appearance of a phytolith did not allow for clear diagnostics of the phytolith (e.g., due to high level of erosion or because it appears largely broken), this was assigned to the group of ‘indeterminate phytoliths’. In all samples of the unprocessed palm leaves (i.e., fresh, dry and herbarized ones), phytoliths whose morphology and preservation allowed clear assignment to any taxonomical plant group, other than Coryphoideae palms, were treated as contaminants, i.e., material not originating from the studied palm-leaf material.

Phytolith identification was performed at the Department of Palynology and Climate Dynamics, University of Göttingen, and partly at the Institute for Wood Science, University of Hamburg. Microphotographs and microslides are kept at the Department of Palynology and Climate Dynamics, University of Göttingen and are available on request from the first author. Plant taxonomy of the phytolith source material followed the guidance of the Angiosperm Phylogeny Group (APG 2016).

Scanning Electron Microscopy (SEM) was applied in order to perform proper identification of the phytoliths registered in the palm leaf material and PLM samples. Small part of leaf tissue of herbarium specimen or a manuscript fragment was cut with an extra-sharp razor blade. The samples were coated in gold dust (BIO-RAD SEM Coating System) and examined in high vacuum by the field emission scanning electron microscope (FE-SEM) Hitachi S520 and FE SEM Quanta FEC 250 at the Institute for Wood Science, University of Hamburg.

We registered 111 unique phytolith types in total, 53 of these were only found in PLM samples. In Borassus flabellifer samples, 104 phytolith types were identified; and in Corypha umbraculifera 89 phytolith types were found ( Figure 2 ). Arecaceae phytolith morphotypes included spheroid psilate, spheroid echinate and echinate symmetrical types, spheroid verrucate symmetrical and spheroid verrucate asymmetrical types, spheroid morphotype with acute projections, spheroid echinate morphotype with small, reduced, or undeveloped spines, spheroid echinate favose type, and aspherical echinate morphotype with sharp-rounded and roundish-triangulate spines, ranging from 9.5 to 14.5 µm in diameter. Both articulated and isolated spheroid phytoliths were grouped here. A complete list of phytolith morphotypes described from all study material and for both palm species is available in Supplementary Material S3 .

In the unprocessed Borassus leaves 40, 42 and 36 types were registered for fresh, dry and herbarized material, respectively. In PLM samples of Borassus, 101 phytolith morphotypes were encountered, of which 23 were registered only in the manuscript material ( Figure 2A ). Phytolith assemblages of unprocessed Borassus leaves were dominated (ca 60% of TPS) by spheroidal echinate bodies of 15-20 µm ( Figures 3A–D ), thus, in the PLM material their contribution to TPS is reduced to ca 50% ( Figure 2B ). Unprocessed Corypha material yielded 48 types in fresh, 39 types in dry and 44 types in herbarized leaves. In the manuscript samples of Corypha, 81 phytolith types were found; 7 types are only registered in the Corypha manuscript samples ( Figure 2C ). spheroidal echinate bodies of 15-20 µm as well as smaller spheroid echinates (<7 µm) were constantly present in all samples (contributing from 59% to 61% altogether to the TPS, Figure 2D ), but were absent from the record of Corypha PLMs. SEM microscopy clearly demonstrated that the small spheroids were a part of the Corypha leaf parenchyma. In addition, on the surface of PLMs of Corypha, spheroidal echinate bodies of a slightly different than Coriphoideae morphology were observed (compare Figures 3F–J, L ). This concerns the number of spines in those that ranged from 10 to 15 (in Coryphoideae number of spines ranges from 17 to 21), and the length of spines ranging between 1.1 µm and 1.2 µm (in Coryphoideae they appeared to be larger: 1.5 μm to 1.8 μm). Spine edges of this morphological subtype are mostly rounded and rarely triangular or spiny as it is demonstrated by Borassus and Corypha spheroids. Ellipsoidal phytoliths with the morphology and morphometry of Coryphoideae were described as well ( Figure 3M ), thus ellipsoids are encountered not that frequently as spheroids. Additionally, in samples of Borassus large (20-25 µm in diameter; Figure 3E ) echinate spheroids were registered 1-2 times in each material type. The aspherical echinate morphotype was only once registered in the fresh leaf material of Corypha; that could be a malformed spheroid.

Other common phytolith types in the unprocessed material of both species were isolated silicified stomata and stomata complexes (aggregated silicified stomata and leaf tissue cells; Figures 4A, B ) of related species. Among the non-diagnostic phytoliths in both Borassus and Corypha leaves, various woody blocky and other blocky as well as hair-like and elongate entire phytoliths were observed (ca 30% of TPS in both species).

Results of phytolith analysis for both Borassus and Corypha with main parameters of the phytolith assemblage diversity, concentrations and contents of phytoliths in the analyzed 200 samples (100 samples of each species) are graphically presented in the Figures 5 , 7 and summarized in Table 3 . Reniform phytoliths ( Figure 3K ) are grouped in diagrams together. Complete diagrams showing all identified types, morphological varieties and druse-shaped inorganic crystals can be found in Supplementary Material S4 for Borassus and in Supplementary Material S5 for Corypha.

Phytoliths from plants other than Borassus and Corypha were regularly found on the surface of fresh, dry, and herbarized palm leaves of these two species. As Table 4 demonstrates, the most common phytolith contaminants of unprocessed leaves were GSSCP, which are small, lightweight, and prone to wind transportation. Virtually all of them (see Tables 3 , 4 ) were eroded, broken, and strongly degraded, with preservation notably different from the phytolith material collected from leaf tissues and PLM material. Presumably, most (if not all) of these degraded phytoliths originate from soil and/or dust.

The total percentage of GSSCP contaminants in all analyzed unprocessed samples remained about the same (ca. 5% for both palm species; Tables 3 , 4 ; diagrams in Figures 5 , 7 ). In the manuscript samples, this amount was approximately twice as high (about 8% in both Borassus and Corypha samples). Other contaminants of unprocessed leaf material with distinctive origins (such as phytoliths of Musaceae or other than GSSCP phytoliths of Poaceae) occurred only randomly and rarely, i.e., 1-3 times per unprocessed material type in both palms. Generally speaking, fresh, dry, and herbarized leaves of Borassus appeared to be more contaminated than leaves of Corypha (compare the numbers of phytolith types per sample in Borassus vs. Corypha in Figures 5 and 7 ).

CONISS clustering ( Figures 5 – 8 ) revealed distinct groups, with a probability of 0.74 for Borassus and 0.85 for Corypha palm leaves, separating PLM samples from unprocessed leaf samples. This was the first and most distinct cluster of samples in all diagrams. Fresh, dry, and herbarium samples were slightly better statistically separated for Corypha (0.61 probability, which is low), but still values of relative abundances of the phytoliths within all unprocessed samples make hardly a difference of 1%. For Borassus, fresh (BF-1 to BF-25) and the first 11 dry leaf samples (BD-1 to BF-11) were grouped together. Herbarium samples (BH-1 to BH-25) were grouped with the remaining 14 dry leaf samples (BD-12 to BF-25).

In the diagram for Corypha, the nature of all three types of unprocessed research material were revealed clearly and statistically significantly (within 2SD): clusters of FC (0.62 probability), DC (0.60 probability), and CH (0.64 probability) were derived. Clustering of the manuscript samples in both palm species was robust and statistically significant. The minimal probability for the separation of unprocessed and manuscript material was 0.59 for Borassus and 0.71 for Corypha. The null hypothesis that no relationships exist between the type of analyzed samples and relative abundance of the phytoliths within the samples was rejected for all Corypha samples at p=0.05. For Borassus, it was rejected for manuscript and unprocessed samples (p=0.05), but accepted for fresh/dry and dry/herbarium samples as these types of Borassus material failed to be distinguished by cluster analysis.

Multivariate analysis performed on all 100 samples of Borassus demonstrated a measure of the goodness of fit equal to 0.87. The first two dimensions of the linear ordination (PCA, Figures 10A–C ) account for 54% and 14% of the total variance of data. For all 100 samples of Corypha R2 = 0.89, and the first two dimensions of PCA ( Figures 10D–F ) taken 57% and 16% of the total data variance. For both palm species classification appeared rather sharp, with the minimal PCA clustering probability being 0.6 (for dry samples of Borassus). Otherwise, the PCA clusters probability varied from almost 0.9 (0.89 for manuscript samples of Corypha and 0.88 for manuscript samples of Borassus) to about 0.73 for other Borassus PCA clusters and about 0.86 for other Corypha PCA clusters, which is comparable to the results of the CONISS clustering described above.

Groups of samples of the unprocessed (BF, BD, BH) and manuscript material (BM) of Borassus are anticorrelated (mean r² = - 0.61 ± 9.3 at p < 0.001; Figure 10C ). For the unprocessed (CF, CD, CH) and manuscript (CM) samples of Corypha this anticorrelation is even stronger, with mean r² = - 0.75 ± 5.3 (at p < 0.001; Figure 10F ). The SD indices demonstrate, that the statistical variability of Borassus samples (taken together) is higher than of Corypha. That allows to conclude that in general Corypha samples are more homogeneous than Borassus samples in terms of their general phytoliths diversity and abundances. The same is also demonstrated by the direct count of phytolith types in the Borassus and Corypha material (compare Figures 2A, C and 8 , 9 ).

For both species, strong correlation (mean r² at p < 0.001 varies from 0.62 to 0.80) of the unprocessed material with spheroid echinate phytoliths, isolated silicified stomata and stomata complexes was revealed. Almost all non-Arecaceae phytoliths demonstrated either weak (from r² = 0.3 to r² = 0.01 at p < 0.001; 0 ≤ r² < 1) correlation to the unprocessed samples of the both palms or an anticorrelation. Instead, non-Arecaceae phytoliths are strongly correlated with the PLM material (r² = 0.51 to r² = 0.9 at p < 0.001). Thus, eroded GSSCP and hair-like phytoliths appeared to be correlated positively with unprocessed samples of Corypha.

It is worth mentioning that some spheroids were anticorrelated (r² ≤ -0.6, p < 0.001; Figures 10A, D ) with the unprocessed samples as well, i.e., spheroid tuberculate phytoliths, spheroid phytoliths with rounded projections, relatively large (18-20 µm) spheroid granulate phytoliths, isolated spheroid psilate phytoliths of a very small (1-1.5 µm) size (the last two are strongly correlated (r² = 0.7 and higher at p < 0.001) with each other and with relatively large decorated ovoids of 10-12 µm in diameter, with smooth-elongate, tracheid, polygonal plate, and long point phytolith types. These phytolith complexes were observed in the same samples of Borassus manuscript samples (BM-3, BM-10, BM-12, BM-18, BM-20, and BM-21; Figures 10A, C , 11 ).

The second distinct phytolith complex was composed of the spikes, ovals, segmented ovals, club-shaped and shapeless phytoliths (only in Borassus), those are very strongly (0.8 ≤ r² < 1, p < 0.001) correlated to each other and occur in the same samples (BM-1, BM-7, BM-10, BM-15, BM-18 and CM-1, CM-5, CM-9, CM-18). In the samples of Corypha, shapeless phytoliths occur in both manuscript and unprocessed samples, thus demonstrating a weak correlation to any specific group of samples together with the non-diagnostic phytoliths (0.2 ≤ r² < 1, p < 0.001; Figures 10D, F , 11 ). The third complex of phytoliths was created by the rondels of morphotype 2, narrow bilobate and double-peaked bodies. They tended to occur together (as in the samples BM-1, BM-2, BM-3, BM-11, BM-13, BM-18 and in CM-1, CM-8, CM-13, CM-18; Figures 10 , 11 ) and were well correlated to each other (0.6 ≤ r² < 1, p < 0.001). cuneiform bulliform phytoliths strongly correlated to the samples (CM-4, CM-5, CM-11, CM-17, CM-19, and BM-15; Figures 10 , 11 ).

Almost all saddle collapsed phytoliths in our record were dark-colored in our record. They were correlated with samples BM-18, BM-20, BM-22 and the majority (r² = 0.64 at p < 0.001) also with CM-1, where this morphotype contributed 5% to its TOS, and that was the maximal for the saddle collapsed phytoliths in this record. Generally, material of both palm species demonstrated a high similarity in the phytolith assemblages among unprocessed samples (i.e., among samples of fresh, dry and herbarized leaves) and high dissimilarity among PLM samples.

Ordination performed on the manuscript samples of Borassus and Corypha taken together ( Figure 11 ) reflected a low correlation (r² = 0.34, p < 0.001) with a very low probability (P<50) indicating that this analysis failed to reflect any clear relationship of the analyzed samples to the palm species used as their writing supports. The samples are mixed, and the only significant group representing the biological nature of the palm leaves is the group of Borassus manuscript samples grouped in the lower-left part of the graph ( Figure 11 ). These Borassus manuscript samples (BM-3, BM-5, BM-8, BM-11, BM-12, BM-13, BM-14, BM-16, BM-17, BM-18, BM-19, BM-21, BM-22, BM-23, BM-25) come from the same geographical area (Tamil Nadu, India; see Table 1 ) and were likely created using similar plants. Burned Cannabis sp. material characterized samples BM-1, BM-7, BM-15, CM-1, CM-5, and CM-18, and neem tree leaf material was evident (>3%) in CM-4, CM-5, CM-11, CM-16, CM-17, CM-18, CM-19, CM-20, and BM-15. All these manuscripts, except CM-11 (Balinese), are originally from Tamil Nadu, India, or Sri Lanka (see Table 1 ). In addition, we were able to determine larger (10-15 µm in diameter, Figure 4M ) rugose spheroids of Canna indica (as described by Piperno and McMichael, 2020) on the surface of PLMs from BM-7 and CM-4 (Tamil Nadu, India), CM-8 (Kerala, India) and BM-18 (Bali, Indonesia).

Most spheroid phytoliths of palms overlap in size, shape, spinule traits, and the number of surficial projections (Witteveen et al., 2022). For isolated (not aggregated) spheroid echinate phytoliths collected from PLM samples, we attempted to separate subfamilies of Coryphoideae and Arecoideae based on their morphometric characteristics. Arecoideae phytoliths, as described by Benvenuto et al. (2015), have a body size of 6-9 µm in diameter with a higher number of spines (10 to 15) compared to Coryphoideae (17 to 21). The length of spines for Arecoideae ranges between 1.1 and 1.2 µm (compared to 1.5 to 1.8 µm for Coryphoideae). Spine edges of this morphological subtype are mostly rounded. Ellipsoidal echinate phytoliths with comparable characteristics were also described, which we assumed could be a variety of the spheroid echinate phytoliths. Phytolith morphotypes of Arecoideae were never observed in the leaf tissue cuts of Borassus or Corypha palms, but only described from the surface of the PLMs and exclusively in an isolated form, suggesting they come from different Arecaceae species than those used as the writing support. Benvenuto et al. (2015) additionally described elongate psilate and elongate phytoliths with fusiform edges, as well as tabular sublobate phytoliths in Arecoideae, which we observed only seldom and did not assign to any specific plant group due to insufficient morphological evidence.

In addition to Borassus and Corypha, another species from Coryphoideae known for PLM production is Phoenix dactylifera, whose leaves are used as manuscript covers (Mahaparata, 1995). These leaves produce a large amount of spheroid echinate phytoliths (Bamford et al., 2006; described as globular echinate in the original publication), which comprise up to 67% of the phytoliths and do not vary much in size (Katz et al., 2010). Isolated spheroid echinate phytoliths described from PLM samples fit this description well, but it is impossible to discriminate them from other Coryphoideae phytoliths of a similar diameter (15-20 µm).

Cocos nucifera is used in PLM production for brushing material rubbed over the surface of the leaf (Bisoi, 1995) and its ash is presumably used for writing, inking, and/or making engraved script visible and more readable (Subramaniam, 1995; Harinarayana, 1995; Nishanthi and Wijayasundara, 2022). This could explain why some Arecaceae phytoliths identified from PLM samples show clear signs of contact with open fire (partially melted and blackish or brownish in appearance).

Additionally, Sharma et al. (2021) mention the use of coconut oil and palm kernel oil (produced from Elaeis guineensis), but oils are unlikely to be a source of palm phytoliths. The use of Areca catechu leaves for polishing manuscripts is indicated by Alahakoon (2012) and Subramaniam (1995). Fenwick et al. (2011) described Areca catechu and C. nucifera phytoliths as spherical to ellipsoidal, with body sizes of approximately 7.68 µm and 9.25 µm, respectively, which aligns with our observations of small Arecaceae phytoliths from PLM surfaces. Fenwick et al. (2011) also note that C. nucifera demonstrates a large percentage of reniform phytoliths (15.6% of their assemblage). We observed some reniform phytoliths ( Figure 3K ) on PLMs characterized by low amounts of Poaceae phytoliths. Considering the frequent use of the grass parts and products, namely straw, bran, and husks as cleaning and brushing material before writing a text (e.g., Sah, 2002; Meher, 2009; Alahakoon, 2012), we assume that in cases where reniform phytoliths are found on the PLM surface and few or no Poaceae phytoliths are present, C. nucifera leaves or fruit shell material were used for polishing.

Reniform phytoliths ( Figure 3K ) were rarely registered in the unprocessed samples (2-3 samples of each material type for Borassus and 3-4 times for Corypha), but their frequency increased about threefold in manuscript samples (15 in Borassus and 17 in Corypha). We assume these phytoliths do not originate from the palm tissues of the studied species, at least not all of them. Conical echinate and conical tabular phytoliths resemble those described by Witteveen et al. (2022) for Bactris sp. and Bactris simplicifrons, respectively. As Bactris are American palms and not native to India, and since Arecaceae conical phytoliths were only seen in herbarium material, it is highly likely that cf. Bactris contamination resulted from storing Borassus and Bactris collection folders together in the same cupboard.

In many cases, it is difficult or impossible to distinguish between Arecaceae and Zingiberaceae phytoliths without explicit knowledge of their exact plant sources (e.g., Tomlinson, 1990; Kealhofer and Piperno, 1998; Benvenuto et al., 2015). Therefore, phytoliths with overlapping morphology, such as large spheroid phytoliths with acute projections (up to 12 µm), spheroid tuberculate phytoliths, spheroid phytoliths with rounded projections, spheroid echinate silica bodies with regularly arranged projections, and spheroid echinate phytoliths with crowded projections and leaf cone phytoliths, were assigned to both families inseparably. Benvenuto et al. (2015) indicate that not only qualitative morphological similarities but also quantitative characters like length and height show strong overlap in phytoliths of species from Marantaceae and Orchidaceae, complicating family differentiation. Other authors note that sphere and spheroid phytoliths with irregular surfaces are seen in Anacardiaceae (Kealhofer and Piperno, 1998), Marantaceae, and Zingiberaceae (Brilhante de Albuquerque et al., 2013). The use of Anacardiaceae in PLM production is unknown. While Marantaceae is common in the New World (Andersson, 1981; Brilhante de Albuquerque et al., 2013), in SE Asia, only Phrynium and Cucurligo are known from northern Thailand forests (Kealhofer and Piperno, 1998).

Orchidaceae phytoliths can also be easily confused with those of Arecaceae and Zingiberaceae (e.g., Kealhofer and Piperno, 1998) as their spheroid phytoliths have comparable size (4-14 µm) and undistinguishable morphology. The presence of Orchidaceae phytolith material in the PLM samples remains unclear. So far, no mention of these plants in PLM production literature was found (Poliakova et al., in preparation). While their ethnobotanical significance could imply usage, more research is needed to confirm this. Benvenuto et al. (2015) note the difficulty in separating Zingiberaceae phytoliths because many genera produce no silica or only small, indistinct phytoliths. We regularly observed small (1-1.5 µm) isolated spheroid psilate phytoliths, but their classification was challenging. Chen and Smith (2013) indicate that Zingiberaceae seed phytoliths do not differ from other vegetative material, though folded, decorated spheres in Zingiberaceae are noted (Kealhofer and Piperno, 1998). Similar morphotypes are observed also in Anacardiaceae that could be the alternative source for these phytoliths in PLM material.

Large multifaceted polyhedrals are known in Zingiber sp. and Curcuma sp (Kealhofer and Piperno, 1998), and both plants are important in PLM production. Zingiber officinale is used as a conserving agent (Sah, 2006; Sahoo and Mohanty, 2007), and turmeric (Curcuma longa) is applied for leaf seasoning (Chakravarti, 1897; Sahoo and Mohanty, 2004; Sah, 2006), coloring with turmeric juice (Wilson and Rice, 2019), and protection (Bisoi, 1995; Padmakumar and Sreekumar, 2003). Multifaceted polyhedrals of Zingiber/Curcuma were observed in all PLM samples from India and Sri Lanka. We also found large decorated ovoids (up to 10-12 µm; Figure 4K ) of Zingiber sp. described by Kealhofer and Piperno (1998), particularly in Sri Lanka manuscripts. Wang et al. (2022) identify smooth-elongate, tracheid, polygonal plate, and long point phytolith types in Zingiberaceae in subtropical Southwest China. These types were described in our PLM samples, supporting the literature that Zingiberaceae were actively used in PLM production. However, these indicators are insufficient to specify particular plant genera or species.

As discussed by Chen and Smith (2013) and Benvenuto et al. (2015), spheroid echinate phytoliths with irregularly arranged projections, spheroid echinate with short and bold projections, large granulate spheroid phytoliths, and spheroid echinate elongate with clustered projections are difficult to attribute to any specific family. Besides Arecaceae and Zingiberaceae, these phytolith types are also described from Bromeliaceae, Strelitziaceae (Tomlinson, 1990; Kealhofer and Piperno, 1998; Benvenuto et al., 2015), as well as Cannaceae (Pearsall and Dinan, 1992; Kealhofer and Piperno, 1998; Brilhante de Albuquerque et al., 2013), and some Cyperaceae (Wallis, 2003). However, we have no clear evidence of all these families being used for PLM production. Additionally, plants from these families (except Zingiberaceae and Bromeliaceae) are not mentioned in the analyzed literature (Poliakova et al., in preparation).

Regarding Bromeliaceae specifically, this plant family is indigenous to South America; there are no bromeliads native to S and SE Asia (Mabberley, 1997; Gouda et al., 2022). The only Bromeliaceae plant historically known in S and SE Asia is the pineapple (Ananas comosus (L.) Merr.), introduced by the Portuguese in 1548 CE. Therefore, all possible identifications of Bromeliaceae phytoliths have to be attributed to Ananas. According to a study by Corrales-Ureña et al. (2018), the bracts and shell of the pineapple consist of spheroid echinate phytoliths 5 to 10 μm in diameter (described as rosette-like silica-based microparticles with an average size of 8.4 ± 2.5 μm, formed by even smaller micro- and nanoparticles). These are morphologically very similar to those described from palms (compare Figures 3C, D with Figure 4L ). Similar phytoliths were also observed in pineapple by Ferreira and de Araujo (2010). Sah (2002) mentions the use of pineapple in Sinhalese PLM practices, noting that leaves (not fruits) were used. We extracted isolated phytoliths of this morphotype from inside the palm leaf tissues of fresh and dry material and additionally described them from the surface of the PLMs as isolated silica bodies. Considering their similar appearance, overlapping morphometrical parameters, and without knowing the exact origin and taxonomic source of these phytoliths, we cannot conclusively determine their exact source. Therefore, we maintain our identification at the level of the family group.

Volcaniform phytoliths (also described as vegetative trough morphotypes; Figures 4C–E ) of diverse varieties, hat-shaped phytoliths, rectangular (or squarish) phytoliths with protuberances, and roundish phytoliths with protuberances were identified as banana phytoliths, as proposed by Mbida Mindzie et al. (2001); Ball et al. (2006); Neumann and Hildebrand (2009), and Chen and Smith (2013). The presence or absence of processes and the arrangement of ridges in tabular seed phytoliths have been suggested as diagnostic features to distinguish between the genera Ensete and Musa (Lentfer, 2009; Perrier et al., 2011). However, due to the limited reference phytolith material available and the state of preservation not always allowing for proper description of the surface texture of the cone and basal parts of the phytoliths, we did not distinguish troughs based on their fine morphology at this pilot phase to avoid overinterpretation. Based on the micro-characteristics of the eight types of volcaniform phytolith morphotypes proposed by Vrydaghs et al. (2009), we attempted to identify phytoliths of edible banana (Musa acuminata Colla) when the preservation state allowed it.

Certain hat-shaped phytoliths with thin bases are also produced by Marantaceae and Lowiaceae (Benvenuto et al., 2015), but they can be well discriminated from the hat-shaped phytoliths of Musaceae. Moreover, these phytoliths are prone to dissolution and may be underrepresented in phytolith assemblages (Benvenuto et al., 2015; own observations). Furthermore, no plants from Marantaceae and Lowiaceae families are known to be used for PLM production (Poliakova et al., in preparation), and we did not observe any similar types in our samples.

In the samples collected from PLMs (and exclusively therein), we observed all typical GSSCP (Twiss et al., 1969; Terrell and Wergin, 1981; ICPN 2.0, 2019; Figure 6 ). Papillate (nipple-shaped) silica bodies, presumably originating from grass bracts (Kaufman et al., 1972; Sangster et al., 1983; Piperno, 2006), culms, and leaves (Brown, 1984; Ball et al., 1993) were observed in manuscript samples from Kerala and Indonesia together with specific Chloridoideae saddles and double Chloridoideae saddles (Cordova, 2023).

Collapsed saddle and bamboo bulliform flabellate phytoliths ( Figures 4F, G ) observed in Indonesia samples, indicated as characteristic of Dendrocalamus giganteus and Dendrocalamus peculiaris (Wang et al., 2022), suggest that Dendrocalamus was probably also used at some stages of PLM production. Other evidence for bamboo (Bambusoideae) includes long saddles (Gu et al., 2016; Li et al., 2014; Tao et al., 2019), observed in some Indian samples. Most of these silica bodies are dark-colored and sometimes melting-deformed at the edges, suggesting possible burning of this particular plant material, likely if Dendrocalamus sp. was used for making and supporting a fire (Elbaum et al., 2003; Parr, 2006). Bamboo stems are a common and inexpensive fire fuel in S and SE Asia. For CM-1, Dendrocalamus sp. phytoliths appeared neither burned nor fire-deformed but mechanically broken, pressed, and/or driven into the manuscript surface. This suggests bamboo material in CM-1 was used for brushing or smoothing the manuscript folios.

The phytolith complex of tall narrow bilobate, saddle, tall saddle, and bilobate bodies observed in the same sample can indicate the use of Bambusoideae, Oryzeae, or Panicoideae grasses (Kealhofer and Piperno, 1998; Dai et al., 2023). However, we suggest these assemblages are more indicative of Oryzeae, serving as proxies for rice, as these types were often found with considerable amounts (10-12% of the total phytolith sum) of bilobate, bulliform, and double-peak phytoliths diagnostic for domestic rice (Gu et al., 2013; Ma et al., 2016). We cannot, however, completely disregard other plant sources.

Determining the use of domestic rice (Oryza sativa L.) and wild rice (Zizania sp.) is culturally and botanically important. Yost and Blinnikov (2011) identified 23 different morphotypes in Zizania palustris L. locally diagnostic of the species in the USA (Minnesota). They indicated one morphotype potentially diagnostic of wild rice (inflorescence morphotype 1). Similar phytoliths were only found in manuscript material from Burma/Myanmar (BM-17, CM-9, and CM-10), along with rondel morphologies, suggesting the use of Zizania sp. rather than Zea mays, as maize is not mentioned for PLM production, whereas rice is regularly referred to (e.g., Sah, 2002; Padmakumar and Sreekumar, 2003). Although we did not find diagnostic phytoliths of Zea mays (e.g., cross phytoliths larger than 21 μm or wavy-top rondels; Iriarte, 2003; Piperno, 2006; Witteveen et al., 2024), the absence of such phytoliths aligns with our assumption concerning rice usage.

Plateaued saddle morphotype suggests the use of Phragmites australis (Cordova, 2023). Elongate entire, i.e., flat rectangular epidermal cell phytoliths, are produced by all grasses (ICPN 2.0, 2019), supporting the hypothesis of active Poaceae use in PLM creation. However, bulliform flabellate phytoliths occur in both Poaceae and Cyperaceae (Esau, 1965; Metcalfe, 1960, Metcalfe, 1971), making it difficult to use them as definitive diagnostic types for grasses. These phytoliths, although present in PLM samples and absent in fresh, dry, or herbarium material, suggest possible use of Poaceae in combination with Cyperaceae at some stages (e.g., burring, boiling, and/or polishing) of PLM production.

Phytolith analysis demonstrates that samples of PLM from Tamil Nadu (India, see Table 1 ) were actively treated with Zingiber/Curcuma (Zingiberaceae) and possibly with pineapple (the only Bromeliaceae in the study regions). Some of Indian manuscripts (See results) were treated with banana leaves (Musaceae) and/or Canna indica. One Borassus manuscript (BM-10), although grouped with Corypha samples in the last ordination run ( Figure 11 ) due to the presence of Cannabis sp. and cf. Azadirachta indica phytoliths, was likely treated with the same plants as well. The last two plants were found in many Corypha manuscript samples ( Figures 7 , 9 ) used for analysis.

Neem tree is a crucial plant in PLM production (Joshi, 1995; Perumal, 2013; Nishanthi and Wijayasundara, 2022). Phytoliths of cf. Azadirachta indica (described as ‘cuneiform bulliform’ by Gasma et al., 2022) were mainly found in the manuscript material of Corypha palm originating from India and Sri Lanka, aligning with the literature on PLM creation and conservation. Neem leaf and burned Cannabis sp. material was evidenced in Tamil, Sri Lanka and Balinese (only in CM-11) manuscripts, where the neem tree is often mentioned as a plant used in PLM production (e.g., Joshi, 1995; Sahoo and Mohanty, 2004; Sah, 2002; Nishanthi and Wijayasundara, 2022) and is still used as a manuscript conservation agent (Subramaniam, 1995) and a ritual plant (Arumugam, 2020). In some parts of Tamil Nadu, neem trees are worshiped as goddesses (Hertzman et al., 2023), and if the manuscript contain any holy text, they can be also treated with neem leaves for ritual purposes. Cannabis sp., an important ritual plant in India and surrounding countries, is evidenced in the PLM samples by the phytolith complex described by Golokhvast et al. (2018). This includes oval, segmented oval, club-shaped, spikes, and shapeless phytoliths (for Borassus), all 20-50 µm in diameter and 20-70 µm in length, as also seen in our samples ( Figures 5 , 7 – 9 , 11 ). Almost all manuscript samples originating from India, especially BM-15 and BM-18 from Tamil Nadu, bear traces of the Cannabis plant on their surface.

Vitex negundo, used for the preparation and conservation of PLMs, is believed to protect the folios from rodents, insects, and mold (Joshi, 1995; Sah, 2006; Sahoo and Mohanty, 2007; Meher, 2009; Sharma, 2018; Wilson and Rice, 2019; Sharma et al., 2021). Common phytolith types for Vitex include abbreviated stellate, polygonal plate (which can also come from Zingiberaceae), plate-elongate, rectangle, woody-block, and hair-like cells (Wang et al., 2022). Although unequivocal identification based on this phytolith complex is not feasible, Vitex can be considered one of potential sources of the mentioned silica bodies on the PLM surfaces.

Microphotographs presented by Kealhofer and Piperno (1998) and our reference material allowed the identification of some decorated spheroids of Dipterocarpaceae (Hopea sp./Shorea sp. type; Figures 4O–Q ), with distinct ornamentation found in Borassus manuscripts BM-5 and BM-9, as well as spheroid folded phytoliths of Mangifera indica morphotype (BM-21; Figures 4H, I ) and pitted, striated phytoliths of Trema (orientalis) epidermis (BM-12; Figure 4N ). However, these single findings do not allow to draw any conclusion on the use of these plants in PLM creation.

A few scalloped large (ca. 40 µm) roundish phytoliths of domesticated Cucurbita sp (Piperno et al., 2002; Piperno, 2006). appeared in two samples of Indian Corypha manuscripts. The Cucurbita sp. is not mentioned in PLM production literature. Furthermore, these well-preserved phytoliths were not covered by patina and cannot be considered part of the PLM production process. There is no possibility of confusing these phytoliths with those of Lithocarpus sp., as the latter forms distinctive, faceted, spheroidal polyhedrals (Kealhofer and Piperno, 1998), which are normally about half the size of Cucurbitaceae phytoliths. It is most likely a modern contamination.

Phytoliths are produced in different plant parts, resulting in variations in shapes and morphotypes (Rapp and Mulholland, 1992; Ball et al., 1993). This phenomenon, termed multiplicity (Rovner, 1971), complicates identifying source plants. Similar morphotypes may be produced by different taxa, not necessarily closely related, leading to redundancy (Rovner, 1971). For example, spheroid echinate morphotypes can be found in palms (Morcote-Ríos et al., 2016; Witteveen et al., 2022; this study), pineapple (Ferreira and de Araujo, 2010; Corrales-Ureña et al., 2018), plants of e.g., Zingiberaceae, Anacardiaceae, Orchidaceae families (Kealhofer and Piperno, 1998; Benvenuto et al., 2015). In addition, phytolith types like blocky, elongate entire, elongate sinulate, tracheary have low taxonomic value (ICPN 2.0, 2019). In order to avoid overinterpretation, a conservative approach with good reference collections and regional studies on phytolith morphology is recommended.

Some plants important in PLM production, such as those from Fabaceae, Malvaceae, and Piperaceae, produce few phytoliths, while others, such as those from Apiaceae and Rutaceae, produce non-diagnostic phytoliths only (Piperno, 2006). Plants like Carica papaya and Capsicum sp. do not produce phytoliths (Piperno, 2006). Lemon grass (Cymbopogon sp.) often reported to be used for seasoning, oiling, conservation, cleaning and increasing attractiveness of the manuscripts (e.g., Sah, 2002; Sahoo and Mohanty, 2007; Alahakoon, 2012) can only be identified in a complex with other Poaceae. For the same reason, the possibility to clearly identify castor beans (Ricinus communis; Sah, 2002), cinnamon (Cinnamonum zeylanicumi; Sah, 2006), black thorn apple (Datura stramonium; Sah, 2002; Perumal, 2013; Wilson and Rice, 2019; Sharma et al., 2021), Faba sp./Vicea faba (Bisoi, 1995; Sah, 2002; Meher, 2009), indigo leaves and bark (Indigofera tinctorial; Sah, 2002; Meher, 2009; Sharma et al., 2021) purely on the basis of phytolith analysis is doubtful.

As it was discussed by Vrydaghs et al. (2016), in applied soil and archaeological phytolith studies, phytoliths not only come from different taxa, but they can be sourced from the different depositional histories (Shillito, 2011). This aspect of the phytoliths taphonomy is also relevant for the studies of PLMs: we may well expect material sedimented on the surface of the PLMs at different times and during the various steps of the preparation process. The accumulation layers can be extremely thin, discontinuous, interrupted, mixed up or even partly removed, e.g., because of cleaning or just because of the active handling.

Random contamination resulting from handling manuscripts, including food stains, finger oils, various ashes, paints, soils, sediments, blood, pollen, spores, bacteria, and other agents, can affect phytolith assemblages on PLM surfaces. However, modern inorganic dust contamination does not significantly influence the results, as we demonstrated in this study.

PLMs are unique and valuable parts of the global cultural heritage. Invasive sampling can be harmful to manuscripts if applied carelessly, and may conflict with conservation and preservation efforts. Developing minimally invasive manuscript material sampling methods is crucial. While the application of phytolith analysis in PLM studies is promising, careful consideration of these challenges and constraints is necessary to ensure accurate and ethical research outcomes.