Exploring an extinct society through the lens of Habitus-Field theory and the Tocharian text corpus

The study seeks to bridge sociology, linguistics (ancient languages), and archaeology by applying Habitus-Field Theory (HFT) to the Tocharian text corpus in order to test whether HFT and associated methods can be used to analyze extinct societies and extinct languages under active philological edition. Tocharian A and B, an extinct Indo-European branch once spoken in oasis city-states of the Tarim Basin (5th–10th century CE), survive only in fragmentary documents from monastic libraries and clerical/secular bookkeeping offices. The corpus reflects an elite perspective (political and economic elites), yet its size and content—together with external historical and archaeological sources—permit reconstruction of religious views, cultural practices, and socio-economic spaces. The paper’s aims are: (1) test applicability of HFT to an extinct society and language; (2) combine close reading, inductive content analysis, and Multiple Correspondence Analysis (MCA) with archaeological metadata (findspots, materiality) to map social structures and spatial differentiation; and (3) provide a blueprint for interdisciplinary analysis of extinct societies using NLP-enhanced approaches.

The theoretical background draws on Pierre Bourdieu’s Habitus-Field Theory. Society is understood as a relational social space structured by the distribution and conversion of different forms of capital (economic, social, cultural, symbolic). Fields (e.g., economy, academe) are arenas where actors compete over capital values, rules, and legitimate strategies; shared investments and tacit beliefs are captured by concepts such as illusio, doxa (orthodoxy/heterodoxy), and nomos. Elites with large capital volumes enter the field of power, imposing classifications and moral orders via symbolic power; symbolic rule denotes internalization of these classifications, and symbolic violence denotes their devaluing effects on dominated groups. Media (texts, images, architecture) manifest positions in the social order. HFT also posits links between physical and social space: residential patterns, resource access, and local tastes reflect social positions, enabling reconstruction of elite distributions through findspots and materiality of texts. This framework guides interpretation of Tocharian fragments as products of political, spiritual, and economic elites and their struggles over social and moral order.

Design: Sequential mixed methods combining close reading, inductive content analysis, and Specific Multiple Correspondence Analysis (MCA). Data: 10,289 fragments in the CETOM project; 5,276 fragments retained for analysis (sufficient size/translation and complete metadata) from 25 regions in the Tarim Basin/Taklamakan. Sources include textual content plus metadata on findspots, linguistic features (Tocharian A/B; genre; verse/prose; variants), and materiality (wood, wall, textile, paper, object). Close reading: Conducted independently on original TA/TB and on English translations to identify topics, genres, social groups, and locations; convergent judgments informed coding. Inductive coding and regex: A category system (from close reading) defined topic lexicons; Python regex identified topic presence per fragment (1/0). English translations were used for regex and MCA input. Metadata integration: Extracted from CETOM (e.g., language variants, material, find region), linked by fragment IDs; regions also categorized (1=Kuca, 2=Turfan Oasis, 3=Sorcuq, 4=southern outskirts). Missing data were excluded to improve MCA performance. MCA: Specific MCA (FactoMineR in R) mapped co-occurrence structures of variables and observations into a low-dimensional Euclidean space. Active variables: linguistic data (language TA/TB; genre literary/non-literary; verse; prose); media/material (wood, wall, textile, paper, object); content topics (monks/monastery; gods/buddha; jewels metaphor; demons/spirits; heaven; hell; royalty; ministers/public servants; settlement; housing types; landscapes; travel; economic terms; signums; householder; worldly goods). Passive variables: regional data (findspot; fragment counts by site; region code). Dimension selection: Elbow criterion on scree plot suggested a 3D solution; reported variances for leading dimensions: Dim1≈18.45%, Dim2≈9.15%, Dim3≈6.06% (subsequent dims ≈5.58%, 5.05%). Biplots aided interpretation.

• Three interpretable MCA dimensions structure the corpus (total ≈29.8% variance for the first three): 1) an elite axis distinguishing spiritual and secular/political content; 2) an economic/profane axis linked to administrative practice and caravan trade control; and 3) a contrast between secular/economic and spiritual/otherworldly poles among elites. - Dimension 1: Religious themes (heaven, demons/spirits, jewels metaphor for Buddha, hell; also monks/monasteries and gods/buddha) co-occur with mentions of worldly elites (householder, worldly goods, military, royalty, ministers/public servants) and urban/settlement terms and journeys. Key centers include Khitai Bazar (Kucha area) and Qigexing (Yanqi), interpreted as texts relating to spiritual and secular elites. - Dimension 2: Characterized by signums (caravan passes), non-literary economic texts, and materiality on wood tablets or walls; associated with sites in peripheral or non-central locales (Miran, Šaldirarŋ, Qizil Qargha, Qizil). Interpreted as an economic dimension reflecting profane administration, taxation, and trade control. - Dimension 3: Fans out from Dimension 1, contrasting householder/worldly goods/economy/military/signum (secular pole) against heaven/landscapes/demons-spirits/jewels-metaphors/settlement (spiritual pole). Nobility clusters with spiritual elites, suggesting alliance; householders (secular economic elites) are positioned as supporting but morally inferior. - Materiality and space mirror social function: sacred/intellectual texts (ink on paper/textile) in centers vs profane/economic texts (wood/walls) in public or peripheral contexts. - Topic and spatial distributions indicate distinct political-spiritual versus economic elite roles, with economic elites present and relevant (especially in trade), yet subordinated symbolically by religious-political elites.

Findings indicate functional differentiation between political-spiritual elites and economically active elites. Religious and political elites exercise symbolic power, defining moral order, virtues/vices, and interpretive authority via writing and canonical texts; economic elites are morally subordinated (symbolic violence), depicted as supporters through donations and logistical support (trade/caravan taxation). Signums and economic terms attest to the practical control of caravan trade as economic capital, which could be converted into symbolic capital through patronage of Buddhist institutions. The material and spatial evidence corroborates textual patterns: central sites and noble materials align with spiritual-political narratives, while peripheral sites and utilitarian materials align with economic-administrative practice. Parallels to medieval Europe are noted, though economic elites appear more visible in the Tocharian corpus. Commoners are largely absent, implying exclusion from the field of power except where public servants or elders appear as legitimated intermediaries. Overall, HFT effectively links textual content, spatial distribution, and materiality to reconstruct the Tocharian social order and elite dynamics.

The study demonstrates that combining HFT, close reading, inductive coding, and Specific MCA with archaeological metadata enables reconstruction of social structures in an extinct society from a fragmentary linguistic corpus. It reveals a tight political–spiritual elite coalition exerting symbolic dominance over secular economic elites, with trade control and donations mediating conversions between economic and symbolic capital. Materiality and findspots are as informative as content for discerning social function and hierarchy. Future research should integrate archaeological accounts with linguistic cues, trace diachronic linguistic and thematic changes to detect symbolic revolutions, and, where available, use historical census or administrative data to map distinctions between trade, civic, political, and religious centers.

• Preservation and discovery are non-random across sites; sources are biased toward elite production and audiences. - Many fragments are undated, limiting dynamic reconstruction; analysis captures a largely static snapshot. - Only elite perspectives are accessible; commoners are largely invisible in the corpus. - The three interpreted dimensions explain about 29.8% of variance—moderate given corpus fragmentation but indicating further latent structures. - Parts of the corpus remain unpublished and were excluded. - The habitus of Tocharian speakers cannot be directly reconstructed from texts alone; the study refrains from ascribing habitus ex post.