The words on the papers
A glossary of the words these documents actually use — leeningsplaats, opstal, erfpacht, kwitrent, morgen and square roods, veldkornetskap, Weeskamer, vendue, d'oude — each illustrated by a verbatim line from the family's own papers.
Every word below is one the district's own papers use, and each quotation is transcribed as the paper has it. Nothing was looked up: the glossary was already in the documents.
The ground itself
leeningsplaats · loan place, loan farm. The oldest form of tenure here. Not ownership: an annual licence to graze a radius, walked rather than surveyed, renewed by paying the recognitie. It is why the eighteenth century leaves almost no maps.
"De helfte van de opstal der leeningsplaats genaamd de Rondabel gelegen aan de Tarca" — the 1818 Rondabel inventory
opstal · the improvements. The house, the kraals, the walls, the trees, everything a man built. On a loan farm the ground was the Company's, so the opstal was what he owned and what he sold. That is why the 1818 inventory values half the opstal at Rd 500 and never mentions the farm: the farm was not his to value.
erfpacht · hereditary lease, the step between a loan place and ownership.
"Erfpacht (hereditary-lease) survey of the loan-place…" — an erfpacht survey of 1816
quitrent · kwitrent, perpetual quitrent, freehold subject to an annual rent. What the 15 April 1841 grants converted the district's loan farms into, and the moment the paper trail really begins.
"originally granted in quitrent to Cornelis Fl[oris Johannes Meyer]" — the Spreeuwfontein diagram of 1898
gedeelte and restant · portion and remainder. When a farm is cut, the pieces are numbered gedeeltes and what is left of the original is the restant. Both words are still the cadastre's. Middelwater is in four pieces today, and an Eskom servitude diagram of 1989 letters them Gedeelte 4, Gedeelte 5 (Vaste Water), Gedeelte 6 (Avondrust) and Restant.
transport · the deed of transfer, and by extension its number. Transport 294 of 20 Nov 1855. A transport number is the single most useful thing to have when ordering from the deeds office.
Measuring it
The chain runs morgen → square roods → hectares, and all three appear on the sheets here.
| morgen | about 0.8565 hectares, originally the ground a team could plough in a morning |
| square rood | 600 to the morgen |
| hectare | from the metric conversion; the 1971 and later diagrams use it |
"7373 morgen 190 sq roods of land situate in the fieldcornetcy of…"
Two traps in reading them.
The first: the decimal comma. A 1930s diagram writes 104.3638 morgen and a 1971 one writes 10 386,1161 hektaar. Read the second with an English eye and it becomes ten million.
The second: a morgen figure is not a hectare figure multiplied. The conversion 0.8565 is close enough to some cadastral ratios to make a wrong number look like a confirmation, and every farm page here carries a caveat because of it.
Who was standing there
veldkornet and veldkornetskap · the field cornet, and his ward. An unpaid local officer, a farmer himself: he walked the boundary with the surveyor, vouched that the beacons had been shown, and took the inventory when somebody died. His ward is the administrative unit these papers are addressed to.
"KROMRIVIER, Gedeelte 1 van die plaas Kopjes Kraal Nr 123, Veldkornetskap TRAKA, Afdeling PRINS ALBERT" — diagram 3281 of 1933
Fourteen of them are named here, between 1838 and 1844 alone.
Weeskamer · the Orphan Chamber, which inventoried the estate of anyone who died leaving minor children. It is the reason the inside of a house in 1802 or 1818 can be read at all: those inventories exist because children needed protecting, not because anybody was recording daily life.
vendue · the public auction at which an estate's movables were sold. Estate papers list vendue buyers and vendue debtors, which is often the only surviving evidence that two families knew each other.
d'oude · the elder. A patronymic disambiguator, and in a district where a name repeats every generation it is doing hard work.
"Pieter Cilliers d'oude" — an Orphan Chamber inventory of 1793
Related: Jansz, Pzn, "As. son", son of Jan, son of Pieter, son of Almero. On a diagram these are not decoration; they are how you tell two men of one name apart, and getting one of them right has settled more than one identification here.
In the house
The 1802 and 1818 inventories are the only places the indoor world appears, and their vocabulary is casks and iron.
| legger | the largest cask, for wine or water · "een oude legger" |
| aam | roughly 150 litres · "een halve aam" |
| anker | a small keg, about 38 litres · "een anker vaatje" |
| uintjes yzer | an iron for digging uintjes, the wild bulbs, out of the veld |
That last one is worth stopping on. It is a purpose-made tool for gathering an indigenous food, listed among a Dutch-speaking household's possessions in 1802, between the spade and the sacks. The knowledge it represents was not theirs first.
"eenhondert en veertig schaapen, zo klein als groot" — a hundred and forty sheep, small and large
On the map
The names on these diagrams are a description of the country, and they repeat because the country does.
kraal · a stock enclosure, and then the place named for it. fontein · a spring, which in the Karoo is the reason a farm exists at all. kloof · a ravine. poort · a gap where a river cuts through a range. Meiringspoort is the one that mattered here. werf · the farmyard, the built centre of a farm, and the thing a family means when it says the farm's name. aanleg · a place laid out or established.
A farm name in this district is usually just water plus a landform: Vaste Water, Groot Sluitel Fontein, Droege Kloof, Honing Nest Poortje. Read that way, the map is a survey of where it was possible to live.
Assembled from the documents and inventories gathered here. No term is included that does not appear on one of them, and every quotation is transcribed as the paper has it, spellings included, which are evidence and not noise.
Sources
Everything this story rests on, with a link to the original wherever one exists. Where a citation is incomplete it says so, an incomplete citation is a research gap, not a formatting problem. If your work is here and the credit is wrong or thin, please tell us and it will be fixed.
- MOOC8/24.27 - Inventory of the estate of Jacobus Oosthuysen d'oude and his widow Dina Carolina Brits, taken 27 April 1802 Cape Archives (KAB), Master of the Orphan Chamber; transcription published by TANAP / Globalise (Huygens KNAW)MOOC8/24.27, {18020427}view it online →
Research notes
[6 Aug 2026] Discovery item L6-07, and the report's own correction was worth heeding: it suggested per-term COUNTS, and those counts would have been of OUR NOTES rather than of the documents. A tally of how often we use a word is not a fact about the corpus, so the page carries terms and quotations and no counts.
Every term was harvested from the model's own documents/inventory rows and every quotation is verbatim - nothing was looked up. The glossary was already in the papers.
The Afrikaans twin is not a translation of this page so much as its natural home: most of the headwords ARE Afrikaans or Dutch already, and per AFRIKAANS.md the document spellings stay exactly as the paper has them. [8 Aug 2026, quality run] NO EVENT LINKS, AND CORRECTLY SO. A glossary: every entry is a term quoted out of a document this project holds. Its dates are the span of the PAPERS it quotes, 1753 to 1971, and it narrates no occasion at all.
[12 Aug 2026] BOLD PASS, both languages. 0 and said it still had lots of bold words and phrases, on a file the voice check had scored 0/3/0 and closed. Rule 4 counts bold carrying a whole SENTENCE and is blind to bolded PHRASES, so a speckled page reads as clean. checks.py now measures DENSITY - bolded words as a share of the page - against the S035 benchmark's 1.1%, plus an en/af asymmetry check, and this story was on that list. Out: dates, extents, survey records, deed and folio numbers, place names and, in S005, whole bolded sentences. STORY-VOICE.md names those exactly, and the benchmark bolds NOTHING in its narrative. NO WORD, FACT, HEDGE, [TO VERIFY], CITATION OR LINK CHANGED - the edit removed emphasis markers. verify_af passing. GLOSSARY, TREATED DIFFERENTLY AND DELIBERATELY: the bold that OPENS an entry is kept, because in a word list the defined term is exactly the scanning anchor STORY-VOICE.md permits; only the emphasis inside the definitions came out, including three whole sentences that survived the first pass because they happened to start a line. 'Weeskamer.' became 'Weeskamer.' with the stop outside the label, per the convention this pass adopted on S009. STILL UNEVEN AND LEFT FOR A DECISION: 11 terms are bold in English and 12 in Afrikaans, and the glossary holds more terms than that - kloof, poort and werf run on unbolded in the English. Either every term is a label or none is; picking which is an editorial call, not a cleanup, so it is flagged rather than guessed.
[12 Aug 2026] EVERY TERM IS NOW A LABEL, closing the question the bold pass flagged rather than guessed. It was 11 bold in English and 12 in Afrikaans while the glossary defined 18, so restant, veldkornetskap, fontein, kloof, poort, werf and aanleg ran on unbolded in the middle of a paragraph and read as prose rather than as entries. Both languages now carry the same 18, the only difference being vendue against vendusie, which is the languages and not the markup. THE TABLE TERMS ARE DELIBERATELY NOT BOLDED - morgen, square rood, hectare, legger, aam, anker, uintjes yzer sit in the first column of a two-column table, which already does the labelling. Bolding them would be the same fault this pass removed from S007, where bare numbers were bolded inside cells a table had already isolated. Density 3.6% en and 4.2% af, under the 5% line, and a glossary is the one shape where that is the standard working rather than failing. No word or definition changed.
[12 Aug 2026] VOICE REWRITE to STORY-VOICE.md, both languages, and THE PAGE'S OWN PROMISE MADE CHECKABLE. en 2/0/0 -> 0/0/0, af 6/0/0 -> 0/0/0 (narration/bold-sentence/punchline). The bold headwords are untouched: a glossary is the one shape where bold as an index is the standard working rather than failing.
THE PROMISE WAS 'Every quotation below is from a document this project holds' AND THE PAGE CITED NOT ONE DOCUMENT. Seven quotations are now attributed and linked: the Rondabel opstal to DOC-0035, the 1818 inventory twice, the erfpacht survey to DOC-0074, the quitrent grant to DOC-0014 (Spreeuwfontein, 1898), the Kromrivier title block to DOC-0247 (diagram 3281/1933), 'Pieter Cilliers d'oude' to DOC-0088 (an Orphan Chamber inventory of 1793), and the 1802 inventory to DOC-0066. A reader can now test the claim instead of taking it.
ONE BLOCKQUOTE WAS NOT ON ANY PAPER. 'Middelwater today: Gedeelte 4, Gedeelte 5 (Vaste Water), Gedeelte 6 (Avondrust), Restant' was set as a transcription and is the description field of DOC-0040, an Eskom servitude diagram of 1989 - our summary of the sheet, not its words. The portion names ARE lettered on that diagram (PAR-000001's notes confirm 'Gedeelte 5 (Vaste Water)'), so nothing is lost: the terms are now given as prose naming the 1989 diagram, and the quotation marks are gone. 5a: a transcription is a promise about what is on the paper.
FLAGGED, NOT FIXED - ONE QUOTATION THE MODEL CANNOT BACK. '7373 morgen 190 sq roods of land situate in the fieldcornetcy of...' matches no document here. The nearest is DOC-0049, Oude Muure No. 121, diagram 1923/1860, which reads '3330 morgen 190 sq roods of CROWN LAND. Bounded N Minies Kraal...' - same roods, different morgen, different wording. LEFT VERBATIM AND UNLINKED, because changing 7373 to 3330 would be manufacturing a transcription and the sheet it came from may simply not be in the model yet. The opening was rewritten anyway and deliberately does not restate the wider promise.
Narration: 'this project's paper trail', 'Two traps this project has fallen into', 'the reason this project can read the inside of a house', 'this project has turned on reading one of them correctly' and the closing 'the model's own document and inventory rows' all recast onto the papers. The warnings they carried are all kept, including that every farm page here carries a caveat about the morgen-to-hectare conversion. verify_af passing.
The research behind this page, kept separate from it: dated working notes, open questions, and what is not yet settled.