1. The question first
Each dossier starts from one sharp question (“Who whispered in Bremer’s ear?”), gathers everything in the archive that answers it, shows the answer in one chart, the evidence in short cards, and the answer’s limits in an explicit section.
2. How we match names
Machine first, then human. We extract names from the archive’s texts and match them against external records by stated rules: first and last name, short forms (Bill/William), alternative Latin spellings of Arabic names (Ayad/Iyad, Zibari/Zebari), and the Arabic forms themselves. Then we read the strongest candidates’ evidence one by one.
A matching name is not a matching person. Common names are down-weighted, surname-only matches are not shown, and relatives are not confused with the name-holder. We show only public officials and companies; we do not name journalists or private individuals.
3. Confidence levels and chart symbols
Established: the fact is written in the document, or two documents state the same role. Likely: a strong inference we reviewed, though no text joins both sides. Doubtful: may be an OCR artefact or a name clash; counted, never named. Open question: the archive alone cannot answer it.
4. Rebuild any chart yourself
Every chart has four buttons: Table view shows the very numbers it draws; CSV downloads them as a file for any spreadsheet; Cite this; Method opens the counting rules on the same page.
Every CSV row carries the archive document ID (such as ORD-091 or DOD-96681122), so you can open it and check the figure in its text. Full data files: Lab questions, The revolving door, Bremer’s memos, Development Fund for Iraq.
5. How to cite
The “Cite this” button under every chart copies a full reference with a link back to that chart. The general form for a whole dossier:
“Dossier title.” Reading Lab, هَمَر: Archive of the Coalition Provisional Authority in Iraq, version 2026.10.22. Accessed [date]. [URL]
The version number 2026.10.22 is the date of the Lab’s latest build; when figures change after review or new data, the version changes, so your citation still points to what you saw.
6. The limits of the archive
The archive is not the CPA. It holds what was published on the CPA’s website (University of North Texas Libraries and the Internet Archive) and what was later released by the Defense Department, Congress and auditors. A document’s absence does not mean it never existed.
Text is machine-extracted. Most Arabic text is OCR; the original document is the reference, and every document page links to its original file.
Counting is not meaning. We measure presence in the papers, not influence or intent; what the text does not say we leave as an “open question”.
7. Privacy
No name from the detainee lists and nothing from the data room appears in the Lab; FARA payment and contact details stay in their locked wing. The names shown are of public officials and companies that appear in published documents.