Source-critical research instrumentPompeii · late Republic to AD 79
The Pompeii Garum Research Atlas
Fish sauce, trade, and evidence under ash.
Navigate the workshops, vessels, inscriptions, biological traces, and scholarly disputes behind Pompeii’s celebrated fish products.
The atlas keeps observation, interpretation, modelling, and uncertainty separate. “Garum Shop” is a modern label—not a recipe result. Its map withholds false precision and never turns associations into routes.
12audited claims
14primary loci
33scholarly works
18geographic records
Quick routes
Choose a research trail
01 · Evidence atlas
What the evidence can—and cannot—support
Each card exposes its evidence, source links, and limiting condition. Filter by epistemic status or research theme.
12 of 12 claimsOpen a card for the audit trail
PGR-C01
Secure evidence
An active AD 79 workplace
公元 79 年仍在运作的作坊
In its final phase, I.12.8 was actively handling, processing, storing, and selling fish products.
在最后一个使用阶段，I.12.8 确实在处理、加工、储存并出售鱼制品。
Inspect evidence trail +
Evidence
Six partly embedded dolia in room 9
Fish bones and scales in vessels and drains
Recently emptied amphorae in room 13
Convergent residue, stratigraphic, and archaeozoological results
Limiting condition
The evidence establishes fish-product activity, not that every preparation was specifically garum rather than liquamen, allec, salsamenta, or an intermediate product.
Wine biomarkers in four tested dolia and lime in a reused amphora permit several operational reconstructions.
四只经检测的陶瓮中有葡萄酒的生物标记，一只再利用的双耳瓶中检出石灰，可容纳数种关于操作流程的复原。
Inspect evidence trail +
Evidence
Tartaric and succinic acids in four dolia
GC-MS, XRPD, carpological, and context evidence
Oil, fruit stones, resin, and lime in reused vessels
Limiting condition
Wine may be an earlier content or an ingredient; lime may derive from construction, cleaning, or acidity control. Absorbed residues need not be final contents.
Sources, sites, and movement—without drawing what is unknown
Switch among Esri topographic, physical, and geographic reference maps; add DARE ancient-place anchors; filter production, processing, consumption, transport, and comparative evidence; inspect every coordinate and its limiting condition.
Mapped now13 evidenced geometries
Withheld5 exact locations
Network ruleNo invented route lines
GazetteerDARE + Pleiades crosswalk
Loading map workspace…
Geographic lines are intentionally absent.Select an evidence association below to frame its endpoints; no intermediate port or itinerary is implied.
Click a relation to frame its source and findspot. A fallback Pompeii settlement halo may appear for I.12.8 because its exact property geometry is withheld.
Coordinate ledger
Every placement exposes its precision
6 shown on the map · 2 filtered records without exact geometry
PointSettlement-level; do not use for room or property placement
HER-CITYHerculaneum
Findspot / consumptionDirect / secure
PointApprox. 500 m for the archaeological findspot cluster
FOS-STGERVAISFos-sur-Mer
Findspot / consumption · Transport contextDirect / secure
PointApprox. 250 m across underwater harbor deposits
SRC-CAL-PELCalabrian–Peloritan
Source regionStrong attribution
Broad interpretive polygonApprox. 100 km; interpretive region, not a site boundary
SRC-W-SICWestern Sicily
Source regionStrong attribution
Broad interpretive polygonApprox. 100 km; interpretive region
SRC-SETR-NCAMEtruria–N. Campania
Source regionStrong attribution
Broad interpretive polygonApprox. 150 km; composite interpretive region
POM-I12-8I.12.8
Processing / packing · ProductionDirect / secure
Exact point withheldUnmapped exact location
FISH-CATCHFish catch sources
Source regionUnresolved
Exact point withheldUnmapped
Ancient places
Digital Atlas of the Roman Empire ↗, CC BY-SA 3.0; fixed DARE IDs and per-record links. Pleiades cross-references retain CC BY 3.0 provenance.
Modern context
Esri basemaps are visual context only. Modern coastlines, relief, roads, names, and boundaries are not evidence for Roman routes, ports, or production.
Deliberate exclusions
Puteoli and Naples appear only as optional DARE context. Ostia, Massalia, the Rhône corridor, and a precise Pompeii harbor are not asserted as network nodes.
Dataset
pompeii-garum-geography-v1 · DARE checked 27 August 2026 · evidence cutoff 26 August 2026.
03 · Contexts
Keep places, phases, and functions distinct
This is a schematic evidence navigator, not a georeferenced reconstruction. Room-level points are deliberately withheld until an authoritative plan is audited.
6 contexts published here
I.12.8 · schematic planselect a context
Other roomsDomestic fabric reused
Active in AD 79
Processing courtyard
Room 9
Six partly embedded dolia, fish remains, drains, and water-management features anchor the strongest processing evidence.
Context control
Room placement is schematic; the interface does not claim georeferenced room coordinates.
Linked claims
Late Republic–early EmpirePossible salting vatsVIII.7 and I.1 · later filled and abandoned
→
Post-AD 62 rebuildingConversion of I.12.8Domestic property adapted to a modest workplace
→
Final phase · AD 79Processing + repackingDolia, amphora pile, fish remains, residues
A bounded supply model
Imports, secondary handling, and local circulation
Source regionCalabrian–Peloritan + Sicilian fish productsMost Dressel 21–22 in the I.12.8 pile
→
Urban nodePompeii · I.12.8Emptying, storage, processing, possible repacking
→
DistributionPrimarily local / regionalOccasional extra-Italian movement ≠ regular export
04 · Texts & labels
Read the words on their objects
Normalized expansions are not fresh autopsies. Audit grades flag restorations, disputed segmentation, and contexts still needing direct object or CIL verification.
14 shown of 14 texts published herethe atlas assesses these and no others
A secureB core secure, note requiredC restored or disputedD direct verification required
Locus / grade
Text + translation
Inference / limit
A/B
AE 1992, 278a
Mosaic urceus, VII.16.12–15
G(ari) f(los) sco(mbri) / Scauri / ex offi(ci)na Scauri
Flower-grade mackerel garum, Scaurus-associated, from Scaurus’ workshop.
SupportsJoins product, mackerel, Scaurus, and an explicit workshop formula.
Does not proveA mosaic advertisement, not a filled vessel; Scauri should not automatically be translated as ownership.
B
AE 1992, 278b
Same mosaic ensemble
liqua(minis) / flos
Flower/first grade of liquamen.
SupportsEstablishes liquamen terminology inside the same display.
Does not proveThe panel does not preserve Scaurus’ name.
A/B
AE 1992, 278d
Same mosaic ensemble
Liquamen / optimum / ex offici[n]a Scauri
Best liquamen from Scaurus’ workshop.
SupportsAssociates Scaurus’ workshop with liquamen as well as garum.
Does not proveOptimum is a promotional grade, not independent quality control.
B
CIL IV 2572
Pompeii; exact property unverified
G(ari) f(los) per se / ex offic(ina) Scaur(i)
Flower-grade garum per se, from Scaurus’ workshop.
SupportsAttests per se and an explicit Scaurus workshop formula.
Does not proveThe label does not define per se; ‘unmixed’ remains an inference.
C
CIL IV 2574a–b
Pompeii; exact property unverified
Dossier normalization: g(ari) f(los) scombri / A. Umbrici Scauri / ex officina Scauri
Mackerel garum associated with A. Umbricius Scaurus, from Scaurus’ workshop.
SupportsBasis for comparing product-level Scauri with ex officina Scauri.
Does not proveEditions differ over segmentation and repainting; this is not a diplomatic text.
A/B
CIL IV 5682
Urceus; exact property unverified
G(ari) f(los) scombri / Scauri
Flower-grade mackerel garum, Scaurus-associated.
SupportsReal-vessel parallel for the mosaic label.
Does not proveScauri alone does not prove production or ownership.
A
CIL IV 5689
Vesuvian context; Abascantus workshop
G(ari) f(los) scombr(i) / optimum / ex officina A. Umbrici Abascanti
Best flower-grade mackerel garum from A. Umbricius Abascantus’ workshop.
SupportsDirect evidence for an Abascantus workshop.
Does not proveShared nomen does not prove Scaurus owned the workshop or determine Abascantus’ status.
B
CIL IV 5690
Pompeii/Vesuvian environs
G(ari) f(los) scombri / Scauri / ex officina Agathopi
Mackerel garum, Scaurus-associated, from Agathopus’ workshop.
SupportsJuxtaposes a Scaurus product designation and Agathopus’ workshop.
Does not proveThe legal or economic relationship is not specified.
B
CIL IV 5711a–c
Urcei; exact property unverified
Liquamen / optimum / A(uli) Umbrici Scauri
Best liquamen of/associated with Aulus Umbricius Scaurus.
Several workshops formed one Scaurus-owned, centrally controlled enterprise.
Named workshop formulas do not state ownership. Lease, contract, dependent management, and independent production remain possible; I.12.8 has no Scaurus urcei.
A Scaurus-associated network of products, workshops, and handlers is attested; ownership and contractual structure are unknown.
02
Market share
Roughly 28–30% of labels indicates market domination.
The denominator is Curtis’ surviving inscribed Vesuvian/Campanian fish-sauce vessels—not all vessels, production, consumption, or sales.
Scaurus names are unusually prominent in the surviving inscribed corpus, but this is not a quantified market share.
03
Distribution
A few Gallic finds and Pliny imply extensive Mediterranean export.
Pompeian Form VI urcei are overwhelmingly local or Campanian. Two Fos-sur-Mer outliers cannot establish a regular route or its scale.
Distribution was primarily local/regional; occasional movement beyond Italy is attested, not regular large-scale export.
04
Product terminology
Garum, liquamen, muria, and allec map neatly to fixed recipes.
The mosaic proves garum and liquamen were distinct labels, not what their biochemical recipes were. Modern reconstructions remain contested.
Retain the Latin terms; distinguish inscriptions, residues, and competing recipe reconstructions.
06 · Terminology
Do not make every fish product “garum”
Ancient labels, biological residues, and reconstructed recipes answer different questions. The Latin terms remain visible where translation would overclaim.
5 product terms published here
Securely say
A Pompeian commercial and literary label; mackerel (scomber) is explicit in several Scaurus-associated texts.
Whether all archaeological fish mixtures called ‘garum’ by modern excavators would have carried this ancient label.
现代发掘者称作 garum 的鱼类混合物，是否在古代都会用这个名称，尚无法确定。
01Fish bonesIdentify taxa and biological profiles
≠
02Chemical residuesIndicate broad materials and vessel histories
≠
03Painted labelsName products, grades, people, and weights
≠
04Recipe textsOffer comparanda, not direct identification
07 · Bibliography
An audited reading room
33 high-value records. Core Pompeii studies are separated from context and methodological comparators; access labels describe the checked landing point.
33 worksCutoff: 26 August 2026
IN PRESS
corein press
Garum in Herculaneum, Pompeii and the Roman Empire
Darío Bernal-Casasola and Daniela Cottica
A verified forthcoming synthesis with no final year, pagination, DOI, or public text at the cutoff.
Full citation
Bernal-Casasola, Darío, and Daniela Cottica. “Garum in Herculaneum, Pompeii and the Roman Empire.” In The Oxford Handbook of Pompeii and Environs. In press.
The most recent completed study devoted specifically to Pompeian garum.
Full citation
Kraus-Rebernig, Claudia Christine. Garum in Pompeji: Untersuchungen zu Produktion und Handel mit Hilfe archäologischer und schriftlicher Quellen. MA thesis, Universität Wien, 2024.
Steven J. R. Ellis, Allison L. C. Emmerson, and Kevin D. Dicus
The definitive stratigraphic publication of earlier possible fish-salting installations.
Full citation
Ellis, Steven J. R., Allison L. C. Emmerson, and Kevin D. Dicus. “Phase 4: Fish-Salting and Other Activities across Insulae VIII.7 and I.1.” In The Porta Stabia Neighborhood at Pompeii, vol. 1, chapter 11. Oxford, 2023.
Confirms fish-product use of the form without proving liquid garum in every vessel.
Full citation
Bernal-Casasola, Darío, et al. “Dressel 21–22 Italic Amphorae for Fish: The Archaeozoological Confirmation from the Garum Shop at Pompeii.” In Roman Amphora Contents, 437–452. Oxford, 2021.
An experimental annual-capacity estimate; the figures are modelled rather than observed.
Full citation
Rodríguez Alcántara, Álvaro, et al. “Analysis of the Productive Capacity of the Dolia Found in the Garum Shop (I 12,8) Using Experimental Archaeology Methods.” In Fecisti Cretaria, 227–230. Rome, 2020.
Desarrollo tecnológico en la recuperación experimental del proceso
Álvaro Rodríguez Alcántara
A technical thesis with substantial use of I.12.8 residues and reconstructed sauces.
Full citation
Rodríguez Alcántara, Álvaro. Desarrollo tecnológico en la recuperación experimental del proceso de elaboración de salsas de pescado romanas. PhD diss., Universidad de Cádiz, 2020.
The gaps below are not generic wish-list items. Each follows from a specific mismatch between evidence, method, and recurring claims.
4 gaps published here
G01
Build a city-wide census
Join workshops, possible vats, amphorae, labels, residues, and chronological phases without treating unlike contexts as equivalent.
Next useful move
Context-aware relational dataset with findspot and phase fields.
G02
Resolve commercial roles
Scaurus, the Popidii, Fortunata, Agathopus, Abascantus, owners, managers, handlers, and customers remain only partially distinguishable.
Next useful move
Object-level label concordance plus property and prosopographical audit.
G03
Test new biomolecular methods
No direct I.12.8 aDNA, palaeoproteomic, or stable-isotope study was located by the literature cutoff.
Next useful move
Chain-of-custody audit and minimally destructive sampling design.
G04
Measure distribution honestly
Market-share and export claims lack a comprehensive, find-context-aware denominator.
Next useful move
Container-form, fabric, inscription, reuse, and discovery-bias model.
Literature-negative result
No direct I.12.8 palaeoproteomics, aDNA, or stable-isotope study was located.
Scope: peer-reviewed and repository literature audited through 26 August 2026. This is not proof of permanent absence; rerun database searches before citing it in future publication.
09 · Method & audit
How this atlas controls claims
The interface is a research navigator, not a substitute for reading the cited scholarship or inspecting the objects and contexts.
01
Inventory claims
Every major statement receives a stable ID and an evidence-status class.
02
Separate witnesses
Findspots, rooms, vessels, labels, residues, texts, models, and later interpretations remain distinct.
03
Attach limits
Every claim states what the supporting evidence does not establish.
04
Link sources
Claims connect to complete citations and stable DOI, handle, repository, or catalogue URLs.
Open citation-audit warnings 8
The archaeological address is I.12.8, not I.12.2.
Curtis’s 1980 AJA item is a meeting abstract, not a full article.
Curtis’s 1984 Archaeology pagination is 58–59 and 74–75.
The 2017 Bernal-Casasola and Cottica chapter ends at p. 252.
The modelled annual capacity must not be reported as observed output.
The 28–30% figure concerns a surviving inscribed corpus, not market share.
Carannante’s one amphora cannot settle the eruption date.
The Oxford Handbook chapter remains in press without final year or pages at the cutoff.
10 · About this build
How it was made, and by whom
Who chose what this atlas covers, who assembled it, and what was checked afterwards.
01
Direction
Wu Ching-Yuan set the subject, the evidence rules, and the shape of the categories: what counts as secure, where a limit has to be stated, which product terms stay in Latin.
02
Assembly
The literature search, the data tables and the interface were built by Codex and Claude working to those rules, with Wu reviewing and correcting at each stage.
03
Curation
Every claim, limit and status class was read and settled by hand. Nothing here is published because a model produced it.
04
Checking
Readings were collated against the EDCS Latium and Campania dataset; where an edition differs, the item carries the other reading.
Release statusResearch build · audited
Bibliographic metadata and interpretive hazards were checked against DOI, publisher, and institutional records. Exact room coordinates were intentionally omitted; several legacy tituli still require direct CIL or object-image autopsy.
Elsewhere on this siteThe rest of the fish-product system
This atlas is the Campanian end: processing, marketing, consumption. The workshops that supplied it are in Production Landscapes, which holds Baelo Claudia, the Bay of Cádiz, Lixus and Tróia as scored dossiers; the four share their Pleiades identifiers with the places mapped here, so the same site resolves to the same point in both. The route between them is the Baetico-Lusitanian garum route in The Roman West.
CollationEDCS · Latium et Campania
Thirteen of the fourteen texts published here have a record in the EDCS province dataset; the fourteenth, Pliny, is literary and has none. Those thirteen correspond to eleven EDCS records, because AE 1992, 278a, b and d are three labels that EDCS carries as a single record. The readings agree except at CIL IV 2572, where EDCS expands the abbreviation G F as factum and this atlas as flos; both are editorial, and the item states which one its translation follows.
The same province dataset holds 166 records whose text carries garum, liquamen, muria, allec or salsamenta as a whole word; 116 distinct texts once repeated labels are counted once, of which 29 name Scaurus, nearly always in the genitive Scauri as the maker’s name on the vessel. EDCS prints these words as editorial expansions rather than in full: G(arum) occurs eighty-one times in this table against a single plain garum, so a search that leaves the brackets standing finds 39 records instead of 166. Those are counts of the corpus. The fourteen texts above are what this atlas assesses, and they are the only witnesses it stands behind.
Collated against the Latium et Campania / Regio I province table of 2022-09-12, 45,162 records, in the file 2022-09-12-EDCS_via_Lat_Epig-prov_LatiumetCampaniaRegioI-45162.tsv. That table is an extract of the Epigraphic Database Clauss-Slaby (manfredclauss.de) made with the LatEpig 2.0 scraper (Ballsun-Stanton and Heřmánková, doi:10.5281/zenodo.12036539), and it is the uncleaned input to the pipeline in sdam-au/EDCS_ETL (Heřmánková, Petra (2022), Scripts v2.0, Zenodo, doi:10.5281/zenodo.7072355). The cleaned all-province dataset built from those tables is published separately as Heřmánková, Petra (2022), EDCS_text_cleaned_2022_09_12 (v2.0) [Data set], Zenodo, doi:10.5281/zenodo.7072337. Neither is redistributed here; where EDCS prints a different reading, that reading is quoted only to state the difference.