1. The period, the register and the texts Mova reads

Early Modern English (EModE) is the English written and spoken from about 1500 to about 1700: from the first printed English Bibles to the Restoration. It is the English of Tyndale, Shakespeare, the King James Bible and Milton. Its grammar is already close to ours, which is exactly what makes it hard for a parser trained on modern text: almost every sentence looks normal, and then one word in five behaves differently.

This chapter fixes the scope of the manual, names the texts that Mova actually reads, and gives the corpus figures that the later chapters lean on.

1.1 Dates and stages

Yang & Eisenstein (2016, Table 3) split the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME) into 1500–1569, 1570–1639 and 1640–1710 and show that a tagger trained on the latest slice loses accuracy steadily as it moves back in time: 93.6 % on 1570–1639 and 87.6 % on 1500–1569 (SVM baseline). The older the text, the harder it is. The texts Mova reads sit at the easy, late end of that scale, but they keep the morphology of 1611.

1.2 The two registers

1. Living EModE (1500–1700): plays, letters, sermons, pamphlets. Spelling varies, you and thou alternate for social and emotional reasons (Abbott §§231–235), -s and -eth coexist, do is optional.
2. The biblical register (1611 to the 20th century): a frozen selection of EModE forms, kept on purpose for solemnity. It keeps thou/ye, -est/-eth, shalt, unto, behold, it came to pass, subjunctive be, but its spelling is modern, and later revisions quietly modernise the strong verbs (§1.4).

Mova's English texts are almost all in the second register. That is good news: the register is regular, consistent and closed. A short list of forms, written once into the lexicon, covers most of it.

1.3 The line of English Bibles behind JPS 1917

The KJV text in our library is the 1769 text from eBible.org (bible-kjv, public domain outside the UK, see the research library guide docs/papers/corpus-religion.md). The Torah that Mova parses is JPS 1917 from eBible.org (torah-jps1917, public domain, Data and model licences).

1.4 JPS 1917 is not the KJV: a measured comparison

JPS 1917 is a 20th-century text written in a 17th-century style. It keeps the person forms but drops the most opaque strong-verb pasts and some function-word habits. The counts below are whole-word matches over the five books of the Torah, KJV (1769 text, brackets of supplied words removed) against JPS 1917 (the 6 105 sentences of torah.tsv), measured 2026-10-17. Both texts have about 156 600 words.

The only its in the KJV Torah is Leviticus 25:5, «That which groweth of its own accord» (a later printers' change: Abbott §228 notes that its «was not used originally in the Authorized Version»). JPS 1917 writes its 163 times. Everywhere the KJV uses his for a thing («the fruit tree yielding fruit after his kind», KJV Gen 1:11), JPS writes its («fruit-tree bearing fruit after its kind», jps:GEN:12, GEN 1:11).

So for Mova, JPS 1917 is EModE morphology with modern lexis: the hard part is the person and number system, not the strong verbs. The 0.3 tales bring back some of the rest (spake, quoth, methinks; chapter 6).

1.5 The texts in this manual and how examples are cited

The Gutenberg books of the tales pool that carry the most EModE forms (counted by sentences with thou):

Titles from the Title: lines of pg<n>.txt. In the whole pool, thou occurs in 1 279 lines, thee 770, art 437 (most of them the noun), whilst 217, wilt 148, shalt 126, quoth 69, spake 48, ere 38, methinks 8.

The tales use EModE as a marked voice: a king, a frog, a fairy, a jinn, a peasant in a translated folk tale speak in thou; the narrator usually does not. Uncle Remus (PG 2306) and the dialect stories are a different variety (19th-century American and Scots dialect: wuz, kin, twa, maun) and are out of the scope of this manual, though they fail at the same gate (chapter 12).

1.6 How this manual treats «correct»

• The reference grammar for EModE itself is Abbott (1870; new ed. 1884), checked against Hejná & Walkden (2022) for the history and against Kulick, Ryant & Santorini (2022) and Yang & Eisenstein (2016) for what parsers do.
• The reference for the annotation is Universal Dependencies 2.18 for English as Mova uses it (en/src/udreg.rs, en/data/ud-registry-en.tsv): lemma of an inflected form is the modern dictionary lemma (shalt → shall, hath → have, thee → thou), the archaic form is marked with Style=Arch, which the registry allows for AUX, PRON, VERB, ADP and ADV.
• Counts are stated with their date. Every count that says «db» is over the 6 065 JPS sentences of parsedb/current.conllu as of 2026-10-17 21:29; every count that says «SLM now» is the small model en-agreed.bin (2026-10-15) run by en annotate built from main at f3e616747, run twice with byte-identical output (2026-10-17).

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/en-x-emode/01-period-and-texts/