Source-backed argument
Mass production without assuming a modern factory
Roman terra sigillata was often produced through layered arrangements: owners or landlords controlled some resources, operating workshops made vessels, shared facilities concentrated risk, and merchants aggregated output.
01
Terms are already an argument
Praedium names landed property. Figlina can name a clay-working property, yard, works, or district. Officina is the operating workshop identity; dominus or domina identifies an owner, especially in the brick evidence. “Production centre” is a modern archaeological category, not a Roman legal person.
The central danger is to read a stamped name as the owner of the whole centre. On terra sigillata it can instead mark productive responsibility, a workshop head, a dependant, or a batch identity. Brick stamps are often more explicit because they can name property, owner, operator, and date.
A second danger is cartographic: an archaeometric reference group is not automatically a workshop coordinate. The atlas now distinguishes secure production zones, broad compositional source regions, candidate centres, and reference-material localities.
02
Control changes along the production chain
Clay extraction, fuel, water, and fixed installations favour estate- or centre-scale control. Forming and decorating are divisible craft activities suited to an officina. Firing reverses the scale again: expensive kilns, shared risk, batch identity, and technical scheduling encourage centre-level coordination.
This division explains how aggregate output could grow by adding operating units around shared services. It also explains why a production centre could behave like one commercial source while remaining internally heterogeneous.
03
A name group is not a family tree
The Roman familia included enslaved people, freedpeople, dependants, and relatives. Freedpeople could adopt a former owner’s nomen, so several matching names may express patronage or manumission rather than biological kinship. P. Cornelius is therefore best treated as a large patronal or naming network whose exact property base and managerial hierarchy remain uncertain. Fülle’s 69-name total accumulates roughly twenty years of attestations; it is not a simultaneous workforce. Cincelli is supported as a production context by mould fragments, while Ponte a Buriano remains a probable production context inferred from a distinct assemblage and a historical kiln report.
The same restraint applies across distance. Ateian products and names at Arezzo, Pisa, Lyon, and La Graufesenque demonstrate dispersion; they do not choose among branch ownership, freedman migration, producer diaspora, or commercial-name continuity.
At Henchir es-Srira, EX OFICINA NAVIGI gives an unusually explicit workshop formula. It still supports only a bounded correlation among a name, an officina, and a chemically defined production centre—not a landowner, legal status, or biological stemma.
04
The kiln is a technological and institutional bottleneck
At La Graufesenque, firing accounts record personal names, vessel forms, and quantities, while failed loads preserve differently stamped vessels fused together. Multiple workshops therefore used common firings. Someone scheduled delivery, allotted capacity, supervised firing, identified lots, and managed losses; the texts do not identify that organizer’s legal status.
Communal firing is among the strongest organizational facts in the dataset. Common ownership is not.
05
One centre can contain several histories
Lezoux contains multiple quarters. The planned conversion of agricultural buildings at Les Ligonnes is compatible with landowner investment and leasing; other quarters developed more incrementally. A single place name can therefore mask different property histories, operator arrangements, and chronologies.
Large basins or kilns likewise prove capacity, capital, and collective demand without naming the investor. M. Perennius is useful precisely because the infrastructure is impressive while its organizational chart is not preserved.
Campanian archaeometry adds a different kind of granularity: secure production at Puteoli, probable production at Cales, and an unresolved Bay of Naples “Produzione A” are technologically separable without being forced into one mapped factory or family network.
North African evidence requires still finer nesting. El Mahrine has four dispersed loci; Bordj el Djerbi and Henchir el Biar are only a few kilometres apart but retain distinguishable fabrics; Sidi Aïch has six pottery mounds whose fabric groups do not map one-to-one onto workshops; and Sidi Rherib remains deliberately unlocated because its name resolves to two candidate areas. These are centre, locus, installation, fabric, and market observations—not interchangeable dots.
Eastern evidence makes the distinction especially visible. Gunneweg, Perlman, and Yellin’s ETS-I chemistry supports an eastern-Cypriot source region but localizes no kiln or named workshop. At Pitane/Çandarlı, by contrast, intensive survey now supplies wasters, kiln material, and firing furniture, so the production zone can be mapped securely.
06
Tools and vessels travel under different kinds of control
Fülle’s Ianu- sequence through La Madeleine, Heiligenberg, and Rheinzabern separates what is observed from what is inferred: no complete moulds are said to have moved, while 132 of 231 Rheinzabern punches were already known at the earlier centres. Individual migration is one explanation; the stamp reading, personal identity, copying, loan, sale, and succession remain separate questions. None requires a corporate branch.
At the other end of the chain, Port-la-Nautique preserves about 300 kg of complete unused sigillata deposited around 50–60 CE: 428 stamps from 90 dies name 53 putative potters. Van Oyen’s reconstruction follows changing units—firing event, transported batch, warehouse stock, and onward cargo—without assuming that the stamped workshop retained commercial control.
Augusta Emerita supplies the complementary warning. Its 595-seal assemblage is exceptionally informative about supplier names and changing punch formulas, yet earlier local-production attributions were rejected. Consumption density maps market reach, not a workshop location, and reconstructed corridors remain hypotheses.
07
“Family company” is useful only after disaggregation
The Domitii offer the clearest hereditary estate network: property passed between generations and stamps expose recurring owner–operator relations. P. Cornelius offers large-scale patronal affiliation. La Graufesenque offers exceptionally strong centre-level coordination. Ateius offers wide connectivity. These are all “large,” but along different axes.
The most defensible model is a layered industrial ecology: estates and resource holders, operating workshops, shared services, mobile specialists, and merchant aggregation.
A modern integrated corporation remains the least secure default. Roman societas could support partnership, but it did not create the durable, separately incorporated enterprise implied by modern company language.