Exemptio in Árpád-age Hungary, 1100–1301: a corpus study

This study asks what exemptio meant in the charters of Árpád-age Hungary and how its use changed. It reads 10,161 charters of the kingdom with Slavonia and Croatia dated 1100–1301, from Wenzel, Smičiklas, Fejér, the Szentpétery–Borsa register and the papal registers, and finds 468 that free someone from a jurisdiction, a due or a service. Exemption was a worldly instrument of royal government: the king and his family grant 63% of these charters, and only five concern a bishop. Within each of the two large Latin editions its share rises after 1205, stays high to 1290 and falls under Andrew III (in Wenzel from 3.7% to 7.5%, then 3.9%). The commonest exemptions take land or people out of a royal castle's lordship (94 charters) or free them from taxes and dues (96); lay grantees mostly receive castle land, churches freedom from dues and from the king's officers. The study argues that exemption reordered who answered to whom rather than freeing anyone from all power, and that single grants extended the freedoms of the Golden Bull of 1222 to those it did not cover. This study is a corpus study: it counts and reads what a large body of edited medieval documents says about one concept, and sets that against modern scholarship. It was prepared with software and AI assistance. The counts, tables, charts and first draft were generated by a program (publisher.py) from the author's corpus, without a language model. The search rules, the concept's introductory paragraphs and the interpretive paragraphs were written with the help of an AI system (Claude, Anthropic), and a random sample of the matches was read in full to measure how many are true instances (section 2). The author read and approved the text before publication. Readings of the Latin marked as preliminary have not yet had a specialist's close reading. The counts depend on what the editions print and how their editors summarised it: editions are uneven by region and decade, modern summaries carry their editors' interpretations, and OCR may miss or misread words, so the numbers show proportions and trends, not totals. The study has not been peer reviewed. Corrections are welcome and will be published as new versions of this record.