Yeshua’s home dialect was Galilean: a Western strain of Palestinian Aramaic. So a fair question about this whole site is: how much of what I’m reading is actually attested in that dialect? The honest answer requires keeping exact-headword evidence separate from root-family evidence. This page exists because the project would rather show both numbers plainly than blend them into a larger claim.
Every content word in the interlinear carries a provenance for its Palestinian-Aramaic row. An exact-headword join and a root-family join answer different questions, so the counts remain separate. Root-family evidence never receives an exact DJPA badge.
The full 5,980-entry Sokoloff DJPA parse is now woven in. It produces 532 accepted corpus joins: 448 exact-headword and 84 root-family. Every candidate went through a two-model screen with orchestrator adjudication; the 155 ambiguous candidates were resolved with 116 promoted to exact, 8 to root-family, and 31 left unjoined. Existing per-entry provenance remains unchanged.
Two honest numbers, two questions. 61.9 percent of content-word occurrences join an exact DJPA dictionary headword (the word is attested in the Palestinian Aramaic lexicon). A stricter 2.2 percent of interlinear tokens carry a distinct attested Galilean word-form shown in the green row; the rest of the Aramaic form you see is the Peshitta, labeled as reconstruction.
The strict number above counts exact word-forms: the only counting that may wear the green badge. Aramaic itself, though, is built from roots, and a fairer picture of how Palestinian the vocabulary is comes from asking: how many words belong to a root family with Palestinian attestation somewhere in it?
of the 26,917 indexed content-word occurrences sit in a root represented by at least one exact-headword DJPA join (17,097 occurrences). Including roots represented by either join class: 18,263 occurrences, or 67.8%.
distinct Aramaic roots in the corpus contain an exact-headword DJPA join (37.7%). Including roots reached through root-family joins: 489 of 1,110, or 44.1%.
Both numbers are honest; they answer different questions. Exact-headword coverage (61.9%) says this corpus entry joins directly to a DJPA headword. Root coverage (63.5%) says this word belongs to a root that contains an exact join. Adding root-family joins broadens the lexical picture, but never turns family evidence into exact proof.
Status, July 2026: the full Lewis-Gibson 1899 Palestinian lectionary has now been transcribed. 51 pages of gospel text passed the agreement gate, contributing 5,768 accepted tokens; the rest of the edition is front matter and apparatus. Disputed and unreadable material remains excluded. Its verses enter these numbers only after each lection is aligned to its verses and reviewed; nothing here yet reflects it.
The Aramaic backbone of this site is the Peshitta: the Syriac gospels, consulted for every token. Syriac is an Eastern branch of Aramaic, and the Peshitta is a translation of an underlying Greek text: it is a witness to the Aramaic substrate, never the original, and this project never promotes it to one.
Because Syriac and Galilean differ lexically: different words, not just different spellings. A blanket Syriac-to-Palestinian orthographic conversion would close the gap dishonestly: converted Peshitta must stay labeled reconstruction, never attestation. The machinery to reconstruct already exists, and each content word remains traceable through the labeled cascade. The gap is not code, it is witnesses.
Every reconstructed form is marked as reconstruction in the interlinear’s styling and provenance. Nothing grey ever presents itself as green.
Rendered output is checked against an explicit list of forbidden overclaims. Text that says more than the evidence fails review and goes back in the queue.
The exact-headword count is not a ceiling; it is a measurement of the witnesses woven in so far. Since this page was first written, the two largest remaining witnesses have both been brought in-house: a full dialect-tag pull of every corpus word, and the transcription of the Palestinian lectionary itself. Weaving them into the numbers is the work now underway; until it lands, the figures above stand unchanged.
A Gospel lectionary in Christian Palestinian Aramaic (the Western Aramaic dialect that is the closest surviving relative of Galilean), published by Agnes Smith Lewis and Margaret Dunlop Gibson and long in the public domain. Its lections cover 1,496 distinct verses of this corpus (Matthew 473, Mark 117, Luke 416, John 490); on every one of those verses, the Palestinian-Aramaic row can move from reconstruction toward attested witness. The manuscript’s script (a distinctive Estrangela variant) defeats conventional OCR, so it was transcribed in-house instead: all 318 pages read by two independent vision passes and accepted only where the two agreed. 51 pages of gospel text passed the gate, yielding 5,768 attested tokens (the rest of the edition is preface and apparatus). The transcription is done and gated line by line, but placing each line on its exact verse proved harder than the printed lection headers suggest: a naive distribution misassigned readings past each lection’s opening, so the per-verse weave is deliberately held for a proper Greek-anchored alignment rather than shown with errors. The reader-facing numbers do not yet reflect it. Getting the verse-and-word placement right, honestly, is the careful step that remains.
The 5,980-entry parse is no longer future work. Its 532 joined corpus entries remain split into 448 exact-headword matches and 84 root-family matches. The two-model screen with orchestrator adjudication resolved 155 ambiguous candidates: 116 were promoted to exact, 8 to root-family, and 31 were left unjoined.
The Comprehensive Aramaic Lexicon’s per-lemma dialect tagging is now woven into the glossary: each word’s card names the Western or Palestinian dialects that attest it (Galilean, Christian Palestinian, Samaritan). Drawn for all 1,845 corpus words (rate-limited queries of our own list, each cited); of the glossary’s content words, 1,171 carry a dialect attestation and 465 are attested in the Galilean corpus itself. It labels and corroborates each word; it never governs the word choice or its tier.
A separate, deferred acquisition, the Kiraz Comparative Edition of the Syriac Gospels (Sinaiticus, Curetonian, Peshitta, Harklean, hand-aligned), would give the project a second independent Syriac witness. That would sharpen the witness column, not the Galilean one: Syriac is not Galilean, and no edition of it changes that. An independent alignment of the Old Syriac is already built (1,401 evidence rows, nearly all Sinaiticus) and deliberately not activated; it stays unscored until that edition arrives.