The fourth track, the first written right to left, and the first that is not an alphabet. Written for someone who reads printed Syriac — or who reads Hebrew or Arabic and can hear the language — and has never read it from a manuscript. Frequencies are real, counted over 2,475 lines / 21,975 words / 7,900 distinct word-forms of the ground truth for Vienna, ÖNB Cod. Syr. 1 (see corpus/sources.yml). Nothing here is invented for illustration.
This is a small codex — 91 × 75 mm, the size of a hand — of the four Gospels, written at Vienna in 1554 by Moses of Mardin, a priest of the Syriac Orthodox Church who had come west from Mesopotamia, and dedicated to the Emperor Ferdinand I.
The reason to read this manuscript rather than any other: it is the printer's copy. ÖNB's own record says so — Vorlage für den bei Caspar Kraft und Michael Zimmermann 1555 in Wien herausgegebenen ersten Druck in syrischer Schrift und Sprache. The book set from these pages, Widmanstetter's Syriac New Testament of 1555, is the first book ever printed in Syriac. So the hand you are learning to read is the hand a compositor read, and the letterforms of every Syriac type since descend, at one remove, from these strokes.
⚠ The ground-truth repository dates the codex 1545. ÖNB dates it 1554, gives Vienna as the place, and ties it to the 1555 edition. 1545 looks like two digits changed places. We follow the library.
(a) It runs right to left. Everything in this trainer that shows you Syriac — the printed line, your input box, the character-by-character comparison — runs the other way. Nothing about reading changes; a habit does.
(b) It is an abjad: the vowels are not written. Twenty-two letters, all consonants. Syriac has a full apparatus of vowel points, and this manuscript does not use it, and neither does the transcription. You supply the vowels from the word, which means you are reading Syriac, not decoding it. This is the ordinary condition of the manuscript tradition, not an austerity chosen for the exercise.
(c) Letters join. Most letters connect to what follows, so a word is one continuous line with a shape of its own. Eight letters — ܐ ܕ ܗ ܘ ܙ ܨ ܪ ܬ — never join to the letter after them, so they break a word into pieces. A gap inside a word is normal and is not a word boundary. This is the single commonest early misreading.
(d) There are no capitals, and no final forms. Hebrew has five final letters; Syriac has none to speak of (final semkath ܤ occurs exactly once in 2,475 lines). One shape per letter, in up to four joined positions.
Three Syriac scripts share these letters:
this one.
They are the same alphabet. If you learn Serto here, you will read the other two with a week of adjustment, not a new alphabet. That is why this site treats Syriac as one script profile with a hand named per manuscript, and not as three.
⚠ The printed line under each plate is a modern typeface (Noto Sans Syriac Western, a Serto face). It shows you what the words are, never what the strokes look like. Read shapes from the ink and only from the ink.
There are no vowels here, but there are dots, and they are not decoration. 675 of the 7,035 word-forms in this bank appear in more than one pointing — that is, the same consonants with the dots placed differently, meaning different things. A dot is as much a letter as a letter is.
Four jobs, in the order you will meet them:
(a) Seyame — two dots over a word, marking the plural. On 1,182 of 2,475 lines. It sits over a letter of the word rather than at a fixed place — most often over a ܪ (455 of the 1,870 in this bank), then ܝ (286), ܡ (185), ܢ (168). ܟܗ̈ܢܐ kāhnē, "priests" (22×) · ܡ̈ܝܐ mayyā, "waters" (15×) · ܟܢ̈ܫܐ kenšē, "crowds" (14×). Without it the same letters are a singular. This is the first thing to train your eye on, because it changes the sense of the sentence and it is easy to miss in a small hand.
(b) The dot that separates homographs. The textbook case is in front of you 445 times:
| written | dots | reads | count |
|---|---|---|---|
ܡ̣ܢ | dot below | men, "from" | 179 |
ܡ̇ܢ | dot above | man, "who" | 49 |
ܡܢ | none | the scribe left it open | 217 |
One set of consonants, two words, told apart by one dot's position and by nothing else in the letters. And the third row is the honest part: half the time this scribe does not point it at all and the sentence has to decide. Expect that. A diplomatic transcription records what is there, so the trainer will ask you for a bare ܡܢ as often as a pointed one.
(c) The dot that gives a verb its tense. ܐܡ̇ܪ with a dot above is āmar, "he says" — the participle — and occurs 167×. ܐܡ̣ܪ with a dot below is emar, "he said" — the perfect — 79×. Add the conjunction and ܘܐܡ̣ܪ "and he said" is one of the commonest words in a Gospel, 155×. When you misplace this dot you change the tense of the narrative.
(d) The dot over the feminine. Over a final ܗ it marks the feminine suffix — her, not his: ܠܗ "to him" (232×) against ܠܗ̇ "to her" (40×).
⭐ 349 lines in the bank carry no mark at all. The bank is ordered easiest first, and those are where you start.
Punctuation is the scribe's, and it is text. . full stop (3,307×) · ܆ (597×) · : (239×) · ܇ (114×) · ܉ (65×) · ܀ end of section (145×). Transcribe what you see; they are not normalised in this ground truth.
Abbreviation is nearly absent. One sign, , on 140 lines — a stroke over a suspended word, most often ܬܠܡܝ̈ܕܘܗܝ for talmīdaw(hy), "his disciples". After the Latin tracks, where a third of the work is expanding brevigraphs, this will feel like a holiday. It is the honest state of a 16th-century Gospel book: this scribe wrote things out.
Gold. The opening of the text is written in gold block letters. Those lines are segmented in the ground truth as a separate zone and are not in this bank — a different letterform, and a reading exercise of its own.
You do not need a Syriac keyboard. Type romanised and the box converts as you go — one key per consonant:
a ܐ b ܒ g ܓ d ܕ h ܗ w ܘ z ܙ H ܚ T ܛ y ܝ k ܟ
l ܠ m ܡ n ܢ s ܣ e ܥ p ܦ c ܨ q ܩ r ܪ x ܫ t ܬPoints come after the letter they sit on, exactly as accents do in the Greek track: " seyame · ^ dot above · _ dot below. So mlk"a gives ܡܠܟ̈ܐ, and am^r gives ܐܡ̇ܪ.
⛔ The mark keys are ^ _ " and deliberately not . , : — a full stop follows a letter 3,307 times in this text, and would have been swallowed as a dot every time.
An early plate, from Matthew's genealogy:
ܪܥܘܬ. ܥܘܒܝܕ ܐܘܠܕ ܠܐܝܫܝ. ܐܝܫܝ ܐܘܠܕ ܠܕܘܝܕ
Right to left: Rʿūt (Ruth), full stop · ʿŪbīd (Obed) awled (begat) l-Īšay (Jesse), full stop · Īšay awled l-Dawīd — Jesse begat David. Note ܐܘܠܕ awled: ܐ and ܘ and ܕ are all non-joining, so the word arrives in pieces and is still one word. Note too that you recognise the names before you have parsed a letter, which is the technique — let the word tell you what the strokes are.
Corrections are welcome and wanted; the track carries a link on every line. There is no Syriacist on this project yet, and the expert seat for this language is empty.
Practise this hand: Syriac — the printer's copy of the first Syriac book