Evidence register

How we know what we show

Evidence register for the Povoado das Mesas do Castelinho

A virtual reconstruction always shows more than an excavation found. This page is the opposite of that: a list of what is documented, what is inferred, what stands on the ground today because modern hands put it back, and what remains unknown. Every element carries its sources, with page numbers, and the reasoning behind it, so that anyone who knows the site can disagree with it in writing.

Where this comes from, and where it does not

A reading of the published research, made from outside the team

Mesas do Castelinho was excavated by UNIARQ, University of Lisbon, between 1988 and 2018, directed by Carlos Fabião and Amílcar Guerra, with Susana Estrela as co-coordinator from 2006. Everything this page states comes from the research that team published, and from the state fieldwork register, which is public.

This page is built from their work, not with them. The excavation team has not reviewed this register, has not approved it, and is not answerable for any reading presented here. Where there is error, the error is ours. The research credit is entirely theirs.

Nothing here comes from the 34 excavation reports deposited with the state archaeology portal, which carry access conditions and were not consulted. What is cited as register are the public fieldwork records of Endovélico, a different document.

The site, in its most defensible form

Where the sources diverge, this is the version none of them contradicts

The figures and dates below are not the ones that circulate most often. They are what you get from setting every published version side by side and choosing the wording that none of them contradicts. Where the team itself moved over time, the range stays in the text instead of disappearing.

Foundation	Late 5th to early 4th century BC. Founded ex novo, on ground not previously occupied. Dated by Attic Castulo cups and Gaditan amphorae; no radiocarbon.
Area	Three and a half to four hectares. Two platforms: A, upper and broadly circular; B, lower, trapezoidal and larger.
The wall	The settlement began without a wall, acquired one within its first decades, and had let it go before the Roman street grid was laid. No destruction horizon at any point.
Iron Age, continuation	Occupation without interruption from the second half of the 4th to the end of the 3rd century BC, on both platforms. In the 2nd century BC the settlement contracts onto Platform B and the upper platform goes out of use, before Rome.
Roman Republican plan	Three north-south streets and blocks, between the end of the 2nd century and the first quarter of the 1st century BC.
Imperial period	Under Augustus the streets were blocked and the large houses subdivided into small compartments. On the upper platform the municipality's visitor circuit places a Roman residential area of the 1st century BC, reusing Iron Age constructions and abandoned in the middle of that century; from then on the upper platform was apparently no longer in use.
Abandonment	Early 2nd century AD. Gradual and progressive, not sudden.
Hiatus	Early 2nd to 9th century AD. A single sherd of Late Phocaean red slip ware, third quarter of the 5th century AD.
Islamic occupation	9th to 11th centuries. A castle with a settlement beside it on the lower platform. Emiral and Caliphal forms present, Almohad forms absent.
Excavation	Between 1988 and 2018, eight sectors (A1 to A4, B1 to B4). Ground-penetrating radar survey in June 2022 and a listed campaign that summer, under the name PROMECA. None of it in the state register.
Listed status	Property of Public Interest, Decree no. 29/90 of 17 July 1990, Annex II. No special protection zone and no non-aedificandi zone.

One word the excavators refuse: town. "A inexistência de indícios de uma arquitectura pública impede-nos de considerar este aglomerado como uma cidade", the absence of any sign of public architecture prevents us from considering this settlement a town (Guerra and Fabião, 2010, p. 475). The orthogonal plan existed; the town did not.

The key, plainly

Two independent axes, and why there are two

The commonest failure in this kind of work is collapsing two different questions into one score: how well evidenced an element is, and how much of it exists on the ground today. They are separate questions and both have to be answered. A wall can be securely reconstructed and be modern work at the same time.

Certainty: how well evidenced it is

Four bands. An honest E4 is compliant; a dishonest E2 is not. No band is better than another, only more or less supported.

Attested

Directly documented at this site, by a source that describes or depicts this very structure or object. The source must carry a page or figure, and cannot be an unverified one.

The Southwest Script stela: 112 × 51 × 9 cm, 82 signs, 80 of identifiable phonetic value, 21 distinct, published by Guerra in 2009 with the measurements and the count.

Inferred on site

Not preserved here, but securely extrapolated from elsewhere on the same site. Requires the source for the analogous feature and the written reasoning.

The section of the lower platform wall: the stratigraphic sequence (houses, fill, outer face) is unambiguous in its order but not in its chronology, so the structure is modelled and the date left open.

Inferred by parallel

Drawn from comparable sites or general period practice, not from this site. The source must name the parallel and the reasoning must say why it applies here.

The density and stature of vegetation around the settlement: the species are attested by charcoal and seeds, the actual look of any given stand is not.

Conjecture

No direct evidential basis; present because something has to be shown. The reasoning must say so plainly, and say what would replace it.

The roofs of the Republican houses. Not one roof survives anywhere on the site, and the two-storey houses with external stone stairs have no known direct antecedent to copy one from.

Physical state: what exists today

Four states a visitor meets on the ground. This is the disclosure duty most often breached by accident, because relaid masonry photographs exactly like ancient masonry.

Conserved in place

It stayed where the excavation found it, neither lifted nor replaced. This is the state that gives a visitor direct contact with ancient construction.

We do not yet have a safe example to give, and it is better to say so. The site has a long conservation history: raffia netting in 1989 and 1992, drainage trenches in 1992, floors relaid with hydraulic lime and geotextile over the great cut in 1994, cement drainage pipes in 1997, and consolidation work carried out by local masons, who still command the dry-stone technique. Much of what stands today has been treated, and this register cannot yet say, element by element, what was left untouched. That is the work the full register has to do.

Returned to original position

Ancient material put back where it stood, by modern hands. The stone is the same; the placing is ours.

The fortification lines of Sector B2, on the western slope of the lower platform: the 2017 campaign record refers to the "reposicionamento das suas linhas fortificadas", the repositioning of its fortification lines.

Rebuilt on the original

Partly or wholly rebuilt on the ancient remains, including consolidation and relaid floors. What you see has two ages at once.

The floors relaid with hydraulic lime in 1994, one of the conservation interventions the state register documents through the 1990s.

Not preserved, or absent

Never preserved, never excavated, or removed to a museum. It is the state of most of what a virtual reconstruction shows.

The stela, today in the Museum of the Southwest Script in Almodôvar; and every roof, none of which exists anywhere on the site.

The register's full vocabulary carries two more values, both real here: REENTERRADO, for what was excavated and deliberately covered again: the great 1986 cut was cleaned and drawn in 1987, then protected with geotextile; and DESTRUIDO, for what existed and was destroyed, which at this site is not a footnote.

Why two axes: the Sector B2 wall

Securely modelled, and repositioned in 2017

The lower platform wall is a double line: the outer wall of the houses backing onto the fortification serves as the inner face, a regular outer face of schist runs parallel to it, and the space between the two is packed with stone thrown in without coursing. Estrela measures it at "mais de 3 metros de espessura", over 3 m thick, against 1 to 2.5 m on the upper platform, and the two platforms were not built to one section, and cannot be modelled as if they were.

On the certainty axis, then, the structure is well evidenced: E2. The sequence is unambiguous; what is unresolved is the chronology; see the next section.

On the physical-state axis, what a visitor sees today on the western slope was put back by modern hands: the 2017 campaign record refers to the "reposicionamento das suas linhas fortificadas situadas no Sector B2". The stone is ancient, the arrangement is from 2017. ANASTILOSE.

Both statements are true at once and neither substitutes for the other. With a single score, one of them would vanish, and it would always be the second, because it is the awkward one.

What is contested

Shown as contested, not quietly resolved

Where different readings carry equal scientific validity, the standard requires both to be presented rather than picking whichever looks better on screen. Where one reading is better supported, that one is endorsed as the main hypothesis and the other recorded as a minority position, with the name of whoever holds it. Both situations appear below, and they are not the same thing.

Did the wall exist from the start?

Two readings of equal validity

Was the settlement born walled, or did it spend a first period without a wall? The answer changes what a reconstruction of the founding moment is allowed to show.

The wall belongs to the founding phase itself, right after the first houses

On the upper platform there was "um momento prévio à construção da fortificação", a moment before the fortification was built, with post-built huts, hearths of reddened clay and storage vessels. The wall then goes up within the same founding phase, late 5th to first half of the 4th century BC: "É um facto que a sua construção se verifica, essencialmente, na sua fase fundacional mais antiga." On the lower platform, the houses first, the wall immediately after.

Estrela 2010, pp. 7, 10-11, 94

A first "povoado cego" moment, unwalled, and only later the fortification

On this reading the settlement began as no more than a "povoado cego", a blind settlement whose defence is the continuous blank rear wall of houses built back to back, which is not a wall at all. The fortification would then be "uma segunda fase em que se reforçaram as características defensivas do conjunto", a second phase in which the defensive character was reinforced. The 2002 campaign record simply says the wall did not exist in the first phase.

Guerra and Fabião 2010, pp. 471-472; 2002 campaign record

Equal validity. Both are presentable; neither is presented as the answer. This is why the corresponding register element carries variants and is itself capped at E2.

“De um ponto de vista estratigráfico, não foi possível determinar se a muralha assim composta foi, de facto, erguida logo na fase inicial da ocupação ou se correspondeu a um momento posterior.”
Guerra and Fabião, 2010, pp. 471-472. Stratigraphically, it was not possible to determine whether the wall so composed was in fact raised in the initial phase of occupation or corresponded to a later moment

All the readings agree on the sequence: no wall, then a wall, then no wall again, and at no point a destruction horizon: "sem quebras nem níveis de destruição", without breaks or destruction levels. They also agree on what the wall was for: not a defensive line but "uma marca na paisagem, um ponto de referência, mais do que um local que se queria inexpugnável", a mark in the landscape, a point of reference, rather than a place meant to be impregnable. Fabião once wrote of the A1-A2 wall that it was so thin it is "quase difícil chamar-lhe 'muralha'", almost hard to call it a wall. What remains undecided is how long the unwalled moment lasted, and whether it counts as a phase of its own.

The terracotta head: toy or cult object?

Two readings of unequal validity

A small terracotta human head, found in 1995 in Sector A1, in secondary position in the collapse and fill of room XIII, so second half of the 4th to the 3rd century BC. What does it represent?

A toy or ornament (main hypothesis)

Realistic features, possibly a child. The only archaeologist to publish on the piece writes: "um brinquedo, ou mesmo um objecto decorativo. Não nos parece que a figura em questão seja a representação de uma qualquer divindade nem mesmo um ex-voto", a toy, or even a decorative object; it does not seem to us that the figure represents any divinity, nor even an ex-voto. The team calls it "Batata", Potato.

Estrela 2010, pp. 78-79

A "cultural synthesis" of Mediterranean and Celtic traits (minority position)

A bald head and shaved face in the Mediterranean manner, with round frontal eyes and an almost absent nose "em conformidade com as iconografias de matriz céltica", in keeping with Celtic-matrix iconography. This reading appears in the municipality's materials and in a press feature that attributes it to unnamed, unquoted archaeologists.

Municipal materials; a press feature dated 2021, with no identified archaeological authorship

Unequal validity, which is why a main hypothesis is endorsed here. Reading A is the published one; reading B has no support in archaeological publication. It is recorded, with its origin, but not as an equal.

“um brinquedo, ou mesmo um objecto decorativo. Não nos parece que a figura em questão seja a representação de uma qualquer divindade nem mesmo um ex-voto”
Estrela, 2010, pp. 78-79

The two readings agree on the find: 1995, Sector A1, secondary position, Iron Age context. They differ on where the piece is displayed: the archaeological publication places it in the Santa Clara-a-Nova museum, the press feature at MESA in Almodôvar.

The foundation date: a range, not a disagreement

The same team writes both ends

Late 5th century BC or first half of the 4th? This is not a dispute between schools: it is one team stating the same thing at different precisions over twenty years.

Late 5th century BC

"nos finais do século V a.C." in the Mérida paper; "Uma fundação ex novo em finais do séc. V a.C." in the 2010 thesis; "finais do século V – inícios do século IV a.C." in 2019. The 2001 register record goes as far as the second half of the 5th century. The dating rests on Attic Castulo cups and Gaditan amphorae.

Guerra and Fabião 2010, p. 459 and p. 477; Estrela 2010, p. 96; Estrela 2019, p. 228

First half of the 4th century BC

"Fundado na 1.ª metade do século IV a.C.", in a 2023 paper co-authored by Fabião, Guerra and Estrela. It is also the date the municipality publishes in its materials.

Tereso and others 2023, p. 240

This is not a scientific disagreement: it is genuine imprecision, and the only honest way to convey it is to carry the range. Writing only one end is choosing for them.

“finais do século V / inícios do IV a.C.”
Guerra and Fabião, 2010, p. 477; in English, late 5th to early 4th century BC

Every version agrees on two things: the foundation is ex novo, "fundação de raiz em local que não era previamente ocupado", a foundation from scratch on ground not previously occupied; and there is no radiocarbon date for this moment. The whole founding chronology rests on imported pottery.

Corrections to the public record

Errors in circulation, and what is published in their place

None of this is anyone's fault. These are copying errors that settled into respectable sources and propagate because the wrong version is easier to repeat than the right one. Several of these statements were on other pages of this very site until this page was written, and were corrected at the same time. Recording them is the service this page performs; scoring points would be no service at all.

In circulation: That the Mesas do Castelinho stela was published in a 2010 article in the Revista Portuguesa de Arqueologia.

The correct position: That article does not exist. The publication of record is Guerra 2009, "Novidades no âmbito da epigrafia pré-romana do Sudoeste hispânico", Palaeohispanica 9, pp. 323-338, open access.

Verified by reading the complete contents of both candidate volumes of the journal, and confirmed across sixteen papers published between 2009 and 2025: every one that cites the stela's publication cites Guerra 2009. Three real 2010 items exist that a compiler could have blended, one of them on amphorae, in the right journal.

In circulation: That the site has Chalcolithic and Late Bronze Age occupation. It is in the property's registered chronology and asserted in the official description.

The correct position: Stray surface finds only. No stratum and no structure of those periods has ever been found. The excavators describe a "fundação de raiz em local que não era previamente ocupado", a foundation from scratch on ground not previously occupied.

The error has an identified origin in a 1975 reference that "a investigação posterior não confirmou", later research did not confirm, and it also has a point of entry into the state record: a 1999 relocation visit, which wrote "As fases mais remotas de ocupação podem remontar ao Calcolítico e Idade do Bronze final". That was a site visit, not an excavation result, and it came two years after the 1997 record had already called that material "descontextualizado".

In circulation: That José Leite de Vasconcelos visited and documented the site in 1897.

The correct position: He was never there. He published a passing reference, obtained indirectly. The 1897 mention does not correspond to a visit.

Stated three times by the excavators themselves: in the 2008 and 2010 papers, and in the 2010 thesis, which writes that Vasconcelos "toma conhecimento indirecto do sítio, sem o ter visitado", came to know of the site indirectly, without having visited it. The 2023 description of the project archive confirms it from the original source.

In circulation: That the stela was found in 2008, and nothing further. Almost every popular account stops here, and some add a day-precise date that no publication gives.

The correct position: The block was spotted "no final da campanha de 2007", at the end of the 2007 campaign, in Rua 1 of Sector B-3, in the context of stratigraphic unit 726, and was not lifted until September 2008. It had been reused face down, in a fill preparing the circulation surface of the street, "num aterro romano republicano, preparativo de um piso de circulação desta via", and not in the paving itself. It is roughly three centuries older than the surface it lay in. Its original location is unknown, and the excavators judged it of little value for dating the settlement.

The publication of record gives the find, the context and the measurements (112 × 51 × 9 cm, 82 signs, 80 of identifiable phonetic value, 21 distinct), but never writes "face down" and gives no date for the stratigraphic unit. That detail, and the date of the pavement, come from the 2008 and 2010 syntheses, not from the epigraphic publication. It is worth knowing too that the state register never mentions the stela: neither the 2007 nor the 2008 record contains the word. In a survey of 28 stelae from excavated contexts, this is the only one ever found "reutilizada en rellenos viarios", reused in road fills.

Also in circulation

Circulates: "86 signs" on the stela.

Published: 82 signs, 80 of identifiable phonetic value, 21 distinct.

Guerra 2009, p. 328. The figure 86 appears in press reports and has no support in the literature.

Circulates: That the destruction of the site was one episode: the machine of late 1986.

Published: There were three. The levelling of the upper platform by the landowner in the late 1950s, the earthworks for planting the olive grove in the 1960s, and the machine of late 1986.

The first is confirmed as an excavation result in the 1991 campaign record; the second is in print, "um olival de plantação relativamente recente (anos 60 do século XX)", and appears in no register record at all. The olive grove, still there, is itself a destruction episode. Treasure-hunting at the site predates 1986, with a "longo historial" recorded from 1960.

Circulates: A "pact of coexistence" between Romans and residents, presented as a documented event.

Published: It is an inference from the absence of violence, and the excavators say so in the first person: "somos levados a concluir que este desmantelamento do sistema defensivo se processou de uma forma pactuada", we are led to conclude that this dismantling of the defensive system proceeded in an agreed manner. The dismantling and the new street grid would be "duas acções de um mesmo programa", two actions of one programme.

Guerra and Fabião 2010, p. 472. Attribute it; do not turn it into a dated event.

Circulates: That excavation is "ongoing", or alternatively that it "ended in 2018", either stated flatly.

Published: Excavation ended with Sector A4 in September 2018. In June 2022 there was non-invasive ground-penetrating radar survey, and the summer of 2022 carries a listed campaign under the name PROMECA. None of it entered the state register, and nobody has published a closure of the project.

The 2018 record says "A escavação do Sector A4 ... foi terminada", the excavation of Sector A4 was finished; a 2023 paper co-authored by all three directors writes, in the past tense, "Entre 1988 e 2018 foi campo de um projeto arqueológico"; UNIARQ's own bulletin documents the radar and the 2022 campaign. Only the directors can close this question.

Circulates: That the settlement was abandoned because the Roman road bypassed it, stated as fact.

Published: It is the excavators' interpretation, argued from the itineraries, and it is cumulative: the Augustan creation of Lusitania, Pax Iulia as the legal seat 70 km away, the pull of the Augustan mines, a road network that bypassed the Caldeirão pass, and above all the choice of Arandis, probably Garvão, as seat of the civitas.

Guerra and Fabião 2010, pp. 479-487. It is not independently documented. Always attribute it.

What we do not know

Declared gaps, and what would close each one

A gap is a result, not a failure. This section exists so that nobody has to work out for themselves where the information stops, and so that whoever holds the missing piece knows exactly where it fits.

The settlement's entrances have never been identified

Six centuries of occupation, a fortified perimeter, and not one gate located. The visitor circuit says as much to visitors: "Não sendo ainda conhecidas as entradas deste povoado fortificado", the entrances to this fortified settlement not yet being known. For a reconstruction this is decisive: there is no honest way to show where people went in.

→ Excavation in the slopes, or the 2022 radar survey, whose results are not published.

The necropolis has never been found

Six hundred years of occupation and no cemetery located. Estrela writes it conditionally: "Se algum dia houver a oportunidade de identificar e estudar as diferentes necrópoles de Mesas do Castelinho", meaning: if there is ever an opportunity to identify and study the different necropoleis of Mesas do Castelinho. The plural is hers.

→ Survey in the surroundings. The absence is itself informative and is recorded as such.

Sectors A4 and B4 are effectively unpublished

B4 was opened in 2013, A4 between 2014 and 2018, five years of excavation, "no lado oriental da plataforma superior, a Sudeste do Sector A1 e a Norte da mina de água", on the eastern side of the upper platform, with at least eleven compartments by 2016. These sectors are essentially absent from accessible publication; the state register records are the only source. They will not be modelled from neighbouring sectors.

→ Publication by the team, or access to the deposited excavation reports.

Whether ore was smelted on site is unknown

Metallurgy is attested and it is domestic: a slag-covered platform, a furnace with post-holes, a Republican forge, all distributed through the residential blocks rather than segregated: "não ultrapassa o nível familiar, doméstico", it does not go beyond the family, domestic level. But the excavators record denser-looking slag suggesting primary ore reduction, and call for more study. The question is open by their own decision.

→ Archaeometallurgical analysis of material already recovered.

Not one roof survives

No roof has been preserved anywhere on the site. The large houses had a second floor reached by an external stone stair, an arrangement the excavators call peculiar and without known direct antecedent in either the local indigenous world or Roman architecture, so there is nowhere to borrow a roof from. Anything a reconstruction puts on top of these walls is declared conjecture.

→ Tile, daub impressions or collapse layers in an unexcavated sector; or the unread context sheets.

The project archive is catalogued and unpublished

Fifty-four boxes at UNIARQ, holding drawings, slides, photographs, contact prints and reports, described in 2023. About 360 item-level records are loaded in the University of Lisbon's archive system, organised by sector, every one showing "Informação não disponível". The site photographs and slides are decaying. A separate public file of the project's contracts and annual scientific reports also exists, unprocessed.

→ Publication of the descriptions, or consultation in person on request.

There are no images of the site before the destruction

The 1950s photographs are the only pre-destruction images and their whereabouts are unknown. The 1986 television report is not in the public index of the broadcaster's archive, searched with the archive's own engine across 25 terms. The only located television record of the site is a 1999 programme.

→ A written request to the archive by date and subject; regional photographic estates.

How much of the site is excavated is not precisely known

The figure everyone repeats, about 10%, comes from a quotation of Carlos Fabião in a press feature that reads as an older text re-dated, and so cannot be dated. It is plausible and it is the director's own; it is not a published measurement.

→ A direct question to the directors, or a figure published in a synthesis.

Element register

Seven records built by hand, in the shape the full register will take

Each element carries an identifier, sector, phase, type, physical state, certainty band, sources with pages and, the most important field, the paradata: the interpretive reasoning, what was rejected and what would change the reading. The seven below were chosen to cover the hard cases: variants of equal validity, variants of unequal validity, an honest conjecture, an anastylosis that has to be disclosed, an absence that is itself information, and the landscape.

The fields map one to one onto the register schema. When the registo/ directory exists, the registo.json generated from the YAML replaces this list without changing the page. Identifiers and anchors are stable: every element is addressable by its id.

E2: Inferred on siteReturned to original positionVariants

Lower platform wall, double line with rubble fill

MC-B2-FERRO3-004

Sector: B2
Phase: Iron Age, foundation (late 5th to first half of the 4th c. BC)
Type: Wall
Locus: Western slope
Dimensions: Over 3 m thick, counting the fill between the two lines (Estrela 2010, pp. 11-12)

Sources

ESTRELA, Susana, pp. 11-12
GUERRA, Amílcar; FABIÃO, Carlos, pp. 471-472
DGPC, Endovélico / Portal do Arqueólogo, ficha do Povoado das Mesas do Castelinho, CNS 4263. Os 34 registos de trabalhos arqueológicos, 1987-2018, lidos na íntegra., 2002 campaign, Results
DGPC, Endovélico / Portal do Arqueólogo, ficha do Povoado das Mesas do Castelinho, CNS 4263. Os 34 registos de trabalhos arqueológicos, 1987-2018, lidos na íntegra., 2017 campaign, Results

Paradata

A double line: the outer wall of the houses backing onto the fortification serves as the inner face, a regular outer face of schist runs parallel to it, and the space between the two is packed with stone thrown in without coursing. The stratigraphic sequence is unambiguous in its order (houses, fill, outer face), but not in its chronology; see variants. Thickness is of the order of 3 m on the lower platform against 1 to 2.5 m on the upper, so the two platforms are not modelled with the same section; Fabião called the A1-A2 wall so thin it is "quase difícil chamar-lhe 'muralha'", almost hard to call it a wall. It is not represented as an imposing structure: Estrela reads it as "uma marca na paisagem, um ponto de referência, mais do que um local que se queria inexpugnável", a mark in the landscape, a point of reference, rather than a place meant to be impregnable. Physical state ANASTILOSE: the 2017 campaign record refers to the "reposicionamento das suas linhas fortificadas situadas no Sector B2", the repositioning of its fortification lines in Sector B2. What a visitor sees today was put back, and has to be marked as such.

Variants

A. Wall raised in the founding phase itself, right after the first housesESTRELA, Susana, pp. 7, 94
B. A blind settlement in a first phase; wall added at a later momentDGPC, Endovélico / Portal do Arqueólogo, ficha do Povoado das Mesas do Castelinho, CNS 4263. Os 34 registos de trabalhos arqueológicos, 1987-2018, lidos na íntegra., 2002 campaign, Results · GUERRA, Amílcar; FABIÃO, Carlos, pp. 471-472

The excavators declare the doubt probably insuperable stratigraphically: "não foi possível determinar se a muralha assim composta foi, de facto, erguida logo na fase inicial da ocupação ou se correspondeu a um momento posterior". The 2002 record says the wall did not exist in the first phase; the 2010 thesis places it inside the founding phase. Equal validity: both presentable, neither presented as the answer. Every reading agrees on the sequence: no wall, a wall, no wall again, with no destruction horizon at any point.

Open questions

•How long did the unwalled moment last, and does it count as a phase of its own?

Record status: draft · Updated: 2026-09-09

E1: AttestedNot preserved, or absent

Southwest Script stela: the object

MC-B3-REP-001

Sector: B3
Phase: Roman Republican (late 2nd to early 1st c. BC)
Type: Object
Locus: Removed from the site; Museum of the Southwest Script, Almodôvar
Dimensions: 112 × 51 × 9 cm (Guerra 2009, pp. 325-328)

Sources

GUERRA, Amílcar, pp. 325-328

Paradata

E1 for the object itself: the publication of record gives measurements, count and reading. Eighty-two signs, 80 of them of identifiable phonetic value, 21 distinct: "esta epígrafe constitui o mais extenso texto deste grupo de escrita até ao momento conhecido", the longest text known in this script group. Database code BEJ.06.13; it has no number in Untermann's corpus, which predates it. Physical state NAO_EXISTENTE because the piece was removed from the site to a museum; anything a reconstruction puts on the ground is a replica and has to say so. What this record does not assert: the date of the inscription itself, which none of the publications read establishes; and the language, which remains unidentified. The 2021 state of the question says the affiliation "seguimos sin conocer", we still do not know it. The reading of the script as Celtic is a published minority position, and Guerra himself moved from sympathy in 2010 to holding, in 2025, that near-complete translation of the texts is methodologically incorrect.

Open questions

•What is the date of the inscription? None of the publications read establishes it.
•How many stelae does MESA hold? The figure in circulation has no museum source.

Record status: draft · Updated: 2026-09-09

E1: AttestedNot preserved, or absent

Deposition of the stela: fill preparing the Rua 1 surface

MC-B3-REP-002

Sector: B3
Phase: Roman Republican (late 2nd to early 1st c. BC)
Type: Deposit
Locus: Rua 1, stratigraphic unit 726

Sources

GUERRA, Amílcar, pp. 325
ESTRELA, Susana, pp. 7
FABIÃO, Carlos; GUERRA, Amílcar, pp. 99
JIMÉNEZ ÁVILA, Javier, pp. 160, 164

Paradata

E1 for the find context, E2 for the orientation. The block was spotted "no final da campanha de 2007", at the end of the 2007 campaign, in Rua 1 of Sector B-3, "no contexto da UE 726", and lifted in September 2008. It had been reused face down, "num aterro romano republicano, preparativo de um piso de circulação desta via", that is, a Roman Republican fill preparing the circulation surface of this street, a fill beneath the street surface, not the paving. It is roughly three centuries older than the surface it lay in; its original location is unknown and the excavators judged it of little value for dating the settlement. Record explicitly what the publication of record does NOT say: it never uses the phrase "face down" and gives no date for unit 726. That detail and the date of the pavement come from the 2008 and 2010 syntheses, not from the epigraphic publication; hence E2 for the orientation. Record too that the state register never mentions the stela: no 2007 or 2008 record contains the word. In a survey of 28 stelae from excavated contexts, this is the only one ever listed as "reutilizada en rellenos viarios", reused in road fills.

Open questions

•Where did the stela come from? Its original location is unknown.
•What exactly is the date of unit 726? The epigraphic publication does not give one.

Record status: draft · Updated: 2026-09-09

E1: AttestedNot preserved, or absentVariants

Terracotta human head ("Batata")

MC-A1-FERRO2-001

Sector: A1
Phase: Iron Age (second half of the 4th to late 3rd c. BC)
Type: Object
Locus: Room XIII, collapse and fill; secondary position

Sources

ESTRELA, Susana, pp. 78-79

Paradata

E1 for the object and the find record: found in 1995 in Sector A1, in secondary position in the collapse and fill of room XIII, Fase II, so second half of the 4th to the early 3rd century BC. Realistic features, possibly a child; the team calls it "Batata", Potato. The published interpretation is explicit and is the main hypothesis: "um brinquedo, ou mesmo um objecto decorativo. Não nos parece que a figura em questão seja a representação de uma qualquer divindade nem mesmo um ex-voto", a toy, or even a decorative object; it does not seem to us that the figure represents any divinity, nor even an ex-voto. The "cultural synthesis" reading of Mediterranean and Celtic traits, circulating in municipal materials and in a press feature that attributes it to unnamed archaeologists, is recorded as a variant of UNEQUAL validity: it has no support in archaeological publication and is not presented as an equal. A divergence still to resolve: the archaeological publication has the piece on display in the Santa Clara-a-Nova museum; the press feature places it at MESA in Almodôvar.

Variants

A. Toy or decorative object (main hypothesis)ESTRELA, Susana, pp. 78-79
B. "Cultural synthesis" of Mediterranean and Celtic traits (minority position)Município de Almodôvar, percurso de visita das Mesas do Castelinho: painéis, brochura e transcrições dos audioguias. · National Geographic Portugal, "Mais de duas décadas de trabalhos arqueológicos em Mesas do Castelinho", com data de página de 23 de agosto de 2021. Lê-se como texto anterior redatado.

Unequal validity, so the main hypothesis is endorsed and the other recorded as a minority position, with its origin. Variant B rests on an unverified source and a municipal one; neither can support an E1.

Open questions

•Where is the piece displayed today: Santa Clara-a-Nova or Almodôvar?

Record status: draft · Updated: 2026-09-09

E4: ConjectureNot preserved, or absent

Roofing of the Roman Republican houses

MC-GERAL-REP-001

Sector: geral
Phase: Roman Republican (late 2nd to early 1st c. BC)
Type: Roofing

Sources

GUERRA, Amílcar; FABIÃO, Carlos, pp. 475
Município de Almodôvar, percurso de visita das Mesas do Castelinho: painéis, brochura e transcrições dos audioguias.

Paradata

Declared conjecture. Not one roof survives anywhere on the site, and this record has no source that describes or depicts a roof at Mesas do Castelinho. The municipal material mentions coverings of "materiais vegetais", plant materials, but that is interpretive material written for visitors, not an excavation result, and is cited as such. Nor is there a parallel to borrow from: the houses of about 30 m² with a second floor reached by an external stone stair are an arrangement the excavators call peculiar and without known direct antecedent in either the local indigenous world or Roman architecture. What would replace this conjecture with evidence: tile, daub impressions or collapse layers in a still-unexcavated sector, or the 148 context sheets of the 2010 thesis, which are downloaded and unread.

Open questions

•Of what material? No excavation source says.
•At what pitch and over what span? No evidence of any kind.

Record status: draft · Updated: 2026-09-09

E1: AttestedNot preserved, or absent

Surrounding landscape: rockrose scrub, holm and cork oak, barley fields

MC-ENV-FERRO3-001

Sector: envolvente
Phase: Iron Age, foundation (late 5th to first half of the 4th c. BC)
Type: Vegetation

Sources

TERESO, João Pedro; VAZ, Filipe Costa; OLIVEIRA, Cláudia; FABIÃO, Carlos; GUERRA, Amílcar; ESTRELA, Susana, pp. 245, 248, 251, 253

Paradata

Unusually well evidenced, from charcoal and seeds in 71 samples across 46 stratigraphic units in sectors A1, B2 and B3. Holm and cork oak charcoal with rockrose dominates every phase: a landscape "bastante antropizada, sendo dominada por estevais, com algumas manchas florestais de azinheira e sobreiro, assim como campos de cereais, em especial de cevada", heavily anthropised, dominated by rockrose scrub with some stands of holm and cork oak, and cereal fields, especially barley. The soils were already impoverished by the 4th century BC, much like today. Barley is the most ubiquitous crop in every phase. Heather charcoal dominates the pit beside the Iron Age smelting furnace in B2 and the Republican forge in B3, while domestic hearths burned oak and rockrose: a deliberate fuel choice for high heat. Vine appears only in the Islamic phase. There is no olive at all in any phase. The olive grove that covers platform B today was planted in the 1960s and is itself a destruction episode, so it cannot be modelled into any ancient phase. The E1 band covers the species present; the density, stature and actual distribution of any given stand are E3 and belong in their own records.

Open questions

•What density and stature did the wooded stands have? The composition is attested, the structure is not.

Record status: draft · Updated: 2026-09-09

E4: ConjectureNot preserved, or absent

Entrances to the fortified settlement: absence

MC-GERAL-FERRO3-900

Sector: geral
Phase: Iron Age, foundation (late 5th to first half of the 4th c. BC)
Type: Absence

Sources

Município de Almodôvar, percurso de visita das Mesas do Castelinho: painéis, brochura e transcrições dos audioguias.

Paradata

An absence record, and the absence is the content. Six centuries of occupation and a fortified perimeter, and not one gate ever identified. The visitor circuit says it to visitors in its own words: "Não sendo ainda conhecidas as entradas deste povoado fortificado", the entrances to this fortified settlement not yet being known. None of the publications read locates an entrance. Consequence for modelling: there is no basis for representing an access, and representing one would be inventing the most visible element of a wall. Until this is resolved, any modelled human figure entering or leaving the settlement is conjecture on top of conjecture. What would close the gap: excavation in the slopes, or the results of the June 2022 ground-penetrating radar survey, which are not published.

Open questions

•Where did people go in? No source says.
•Did the 2022 radar see anything? The results are not published.

Record status: draft · Updated: 2026-09-09

Standards, revision and errata

What this page holds itself to

An evidence register is not an editorial courtesy: it is what two international standards of virtual archaeology require of any reconstruction presented in public. This page exists to discharge those standards' publication duty: to make the metadata and paradata of everything modelled clearly, concisely and easily available.

Seville Principles

International Principles of Virtual Archaeology, ratified by the 19th ICOMOS General Assembly, New Delhi, December 2017.

Authenticity, historical rigour and scientific transparency. This is the standard that requires the real to be distinguished from the reconstructed, and the paradata to be published.

Trilingual text (English, Spanish and French) published by ICOMOS Spain.

London Charter

The London Charter for the Computer-Based Visualisation of Cultural Heritage, version 2.1, February 2009.

The theoretical framework the Seville Principles adopt as their own. Principles 3 (Research Sources) and 4 (Documentation) are the ones this register discharges; the term paradata is the Charter's.

The Charter has an official Portuguese translation, by Botelho, Dias, Madeira and de Almeida, of November 2014, linked from the same site. "Carta de Londres" is therefore an established Portuguese term, not a rendering of our own.

The clauses this page discharges

Seville 4	It should always be possible to distinguish what is real, genuine or authentic from what is not.
Seville 4.1	Where alternative interpretations have equal scientific validity, there is a commitment to presenting them all. Only where that equality does not exist is the main hypothesis endorsed.
Seville 4.2	Reconstructions must make explicit the different levels of accuracy on which they are based.
Seville 4.3	Distinguish clearly among remains conserved in situ, remains returned to their original position, areas rebuilt on the original remains, and areas virtually restored or reconstructed.
Seville 5.2	Show all phases, including periods of decline, and never only the time of splendour, nor an idyllic image of seemingly newly-constructed buildings.
Seville 5.3	Landscape and context are as important as the ruin itself. Lifeless cities and dead landscapes are historical falsehoods.
Seville 7.1 and 7.3	Gather and present transparently the entire work process; the incorporation of metadata and paradata is crucial and should be clear, concise and easily available.

Source list

The verification mark says how far each item was actually checked, and is copied from the research record, never invented. An unverified source cannot support an E1.

FABIÃO, Carlos; GUERRA, Amílcar (2008), "Mesas do Castelinho (Almodôvar), um projecto com vinte anos", Al-Madan, II série, 16, pp. 92-105.
fabiao_guerra_2008 · publication · read in full
GUERRA, Amílcar; FABIÃO, Carlos (2010), "Mesas do Castelinho (Almodôvar): um exemplo de urbanismo falhado no sul da Lusitânia", VII Table Ronde sur la Lusitanie Romaine (Toulouse 2007), Mérida, pp. 459-488.
guerra_fabiao_2010_merida · publication · read, quotations checked
FABIÃO, Carlos; GUERRA, Amílcar (2010), "Mesas do Castelinho (Almodôvar): a case of a failed Roman town in southern Lusitania", in Corsi & Vermeulen (eds), Changing Landscapes, Bologna, Ante Quem, pp. 325-346.
fabiao_guerra_2010_changing · publication · read in full
GUERRA, Amílcar; FABIÃO, Carlos (2001), "Mesas do Castelinho, Almodôvar: uma fortificação rural islâmica do Baixo Alentejo", Mil Anos de Fortificações na Península Ibérica e no Magreb, Lisboa, Colibri / CM Palmela, pp. 171-176.
guerra_fabiao_2001_islamica · publication · read as page images
ESTRELA, Susana (2010), Os níveis fundacionais da Idade do Ferro de Mesas do Castelinho (Almodôvar): os contextos arqueológicos na (re)construção do povoado, dissertação de mestrado, Universidade de Lisboa, orientada por Carlos Fabião.
estrela_2010_tese · publication · read, quotations checked
ESTRELA, Susana (2019), "Adornos, espaço e tempo: as contas de colar em Mesas do Castelinho", digitAR 6, pp. 227-253.
estrela_2019_digitar · publication · read, quotations checked
GUERRA, Amílcar (2009), "Novidades no âmbito da epigrafia pré-romana do Sudoeste hispânico", Palaeohispanica 9, pp. 323-338.
guerra_2009_palaeohispanica · publication · read, quotations checked
JIMÉNEZ ÁVILA, Javier (2021), "El contexto arqueológico de la escritura paleohispánica del Suroeste peninsular", Palaeohispanica 21, pp. 149-188.
jimenez_avila_2021 · publication · read, quotations checked
LUJÁN, Eugenio R. (2021), "La lengua de las inscripciones del Sudoeste: estado de la cuestión", Palaeohispanica 21, pp. 189-217.
lujan_2021 · publication · read, quotations checked
VIEGAS, Catarina (2019), "Terra sigillata trade in Mesas do Castelinho (Almodôvar-Portugal)", Spal 28(1), pp. 97-129.
viegas_2019_spal · publication · read, quotations checked
PARREIRA, Jorge (2009), As ânforas romanas de Mesas do Castelinho, dissertação de mestrado, Universidade de Lisboa, orientada por Carlos Fabião.
parreira_2009_tese · publication · read, quotations checked
TERESO, João Pedro; VAZ, Filipe Costa; OLIVEIRA, Cláudia; FABIÃO, Carlos; GUERRA, Amílcar; ESTRELA, Susana (2023), "Agricultura e paisagem em torno do povoado de Mesas do Castelinho (Almodôvar)", in Amanhar a Terra. Arqueologia da Agricultura, Município de Palmela, pp. 239-255.
tereso_et_al_2023 · publication · read, quotations checked
SECA, Matilde (2023), "Organização e Representação da Informação do Projeto 'Mesas do Castelinho' no Centro de Arqueologia da Universidade de Lisboa", Boletim do Arquivo da Universidade de Coimbra 36(1), pp. 121-139.
seca_2023 · archive · read in full
DGPC, Endovélico / Portal do Arqueólogo, ficha do Povoado das Mesas do Castelinho, CNS 4263. Os 34 registos de trabalhos arqueológicos, 1987-2018, lidos na íntegra.
registo_endovelico · register · institutional record, read at source
DGPC, ficha de imóvel 72912, incluindo a Nota Histórico-Artística assinada [AMartins]. Bibliografia parada em meados dos anos 1990.
dgpc_imovel · register · institutional record, read at source
Decreto n.º 29/90, de 17 de Julho, Presidência do Conselho de Ministros, Diário da República I Série n.º 163, Anexo II, p. 2973.
decreto_29_90 · register · institutional record, read at source
uniarq digital, boletim do Centro de Arqueologia da Universidade de Lisboa, n.os 65 e 66 (2022) e 88 (2024).
uniarq_boletim · archive · read in full
Município de Almodôvar, percurso de visita das Mesas do Castelinho: painéis, brochura e transcrições dos audioguias.
cm_almodovar_audioguia · municipal · read in full
National Geographic Portugal, "Mais de duas décadas de trabalhos arqueológicos em Mesas do Castelinho", com data de página de 23 de agosto de 2021. Lê-se como texto anterior redatado.
natgeo_2021 · press · unverified

Last revised: 9 September 2026

Errata

If you know the site, dug on it, or published about it, and you see an error here (a misattributed reading, a wrong page, a hypothesis presented with more confidence than it has earned), the correction is welcome and will be made with credit. This register was written so that it could be disagreed with in writing.

Get in touch

Back to Mesas do Castelinho