5. Auxiliaries and modals

The auxiliaries carry tense, mood and the speech act of the biblical sentence: the commandments are thou shalt, the promises I will, the purpose clauses that thou mayest. They are also where Mova's pipeline broke first: on 2026-10-17 the Torah cycle lost 1 087 distinct sentences to one gate message, «AUX with lemma 'shalt' is not in the list of auxiliaries» and its siblings (chapter 12). This chapter gives the forms, their meanings, their UD analysis, and the measured state.

5.1 Forms

Spelling variants in the tales: may'st, could'st, would'st, should'st, hadst, canst, shan't, 'twould. Source for the paradigm: Abbott 1884, §§298–331 (Be, Were, Do, Can, May, Must, Shall, Will, Should, Would); Hejná & Walkden 2022, §4.4.1 box «The modals».

5.2 Meanings that differ from modern English

shall / shalt

In the biblical register shall is mostly deontic (an order, a law) in the 2nd and 3rd person and predictive in the 1st: «Thou shalt not murder» (jps:EXO:566, EXO 20.13) is a prohibition, not a forecast. Counts in JPS 1917 (2026-10-17): thou shalt 710, shalt not 161, shalt thou 160, shalt surely 17, ye shall 388, shall not 198, shall be 717.

Abbott §317: «Shall. You shall. He shall. On the other hand shall, …» — in Shakespeare shall with the 2nd and 3rd persons keeps its old meaning of obligation or necessity; §318: with the 1st person it states the speaker's future without volition.

For Mova's pragmatics, thou shalt + verb is a directive speech act (act=order or a law), even though the mood is indicative. prag/src/feats.rs takes the modal from the form (norm_word), so shalt and shall are two different feature values, and shalt, wilt, mayest, canst, shouldest each start from zero evidence. Chapter 13 (P2) proposes mapping them to the modern modal.

will / wilt

will keeps its old lexical meaning «want, intend»: «Wilt thou go with this man?» «I will go.» (jps:GEN:683, GEN 24.58) — Rebekah is asked whether she is willing, not what will happen. I will 297 and will I 50 in JPS are mostly promises and decisions. Would = «wish»: «Would that we had died by the hand of the LORD» (jps:EXO:448, EXO 16.3), would that 6. Abbott §329: «Would for will, wish, require»; §330: «Would often means "liked", "was accustomed"».

may / mayest / might

may is mostly the purpose subjunctive substitute: that X may 164, that X mayest 28, that X might 25: «that thou mayest know that there is none like unto the LORD our God» (jps:EXO:212, EXO 8.10 in the numbering of torah.tsv.verses). Also permission and wish («May God give thee…»). Abbott §307: may originally meant «to be able».

can / canst

Ability; also «know how to»: «whereof thou canst not be healed» (jps:DEU:764, DEU 28.27). can 17, canst 4, could 19 in JPS.

should / shouldest

Obligation and the conditional after lest: «lest thou shouldest say: I have made Abram rich» (jps:GEN:375, GEN 14.23); a hypothetical question: «shouldest thou therefore serve me for nought?» (jps:GEN:866, GEN 29.15). Abbott §§322–326.

must, ought, need, dare

must 21, ought 2 in JPS («deeds unto me that ought not to be done», jps:GEN:531, GEN 20.9). need and dare behave as modals without -s and without do: «he need not», «durst not».

do

do in the biblical register is either the lexical verb («Noah did according unto all», jps:GEN:176) or an auxiliary that is often not required by grammar: «God doth know» (jps:GEN:65, GEN 3.5), «did eat their portion» (jps:GEN:1563, GEN 47.22), «God did prove Abraham» (jps:GEN:576, GEN 22.1). Chapter 7 treats do in full.

be as a perfect auxiliary

With verbs of motion and change, EModE forms the perfect with be, not have:

Abbott §295: «With some few intransitive verbs, mostly of motion, be is used» for the perfect. These are not passives: come is intransitive. In UD the be is aux and the participle has Tense=Past|VerbForm=Part without Voice=Pass. The db is mixed: of the patterns found, are come is aux + no voice 10 times but aux + Voice=Pass 5 times; was come/were come are aux:pass + Voice=Pass 2 + 3 times; be waxen is a copula with an adjective (4) — acceptable for waxen (= grown), but then its lemma should be wax. A rule: be + past participle of an intransitive verb of motion or change (come, go, rise, wax, become, depart, enter, return, fall, grow) → aux, no Voice=Pass.

have as a full verb and had rather

have + infinitive with to = must («I have to tell thee») is rare in JPS. had rather, had liefer, were better (preference) occur in Shakespeare and the tales.

5.3 Word order with auxiliaries

• Inversion in questions: «Hast thou eaten of the tree?» (jps:GEN:72, GEN 3.11); «Art thou my very son Esau?» (jps:GEN:798, GEN 27.24); «Wilt thou go with this man?» (jps:GEN:683).
• Inversion after a fronted adverbial or object: «with rooms shalt thou make the ark» (jps:GEN:163, GEN 6.14); «therefore shall his days be a hundred and twenty years» (jps:GEN:151, GEN 6.3).
• Negation after the auxiliary or after the subject: «thou shalt not eat of it» (jps:GEN:51); «hast thou not withheld» type; with inversion not follows the subject: «knowest thou not» (jps:EXO:277).
• Stranded auxiliary: «'Wilt thou go with this man?' And she said: 'I will go.'» — modern ellipsis would answer «I will»; JPS repeats the verb.

5.4 UD analysis

The registry's list of auxiliary lemmas (en/data/ud-registry-en.tsv) is: be, can, could, dare, do, get, have, may, might, must, need, ought, shall, should, will, would. Every archaic form above lemmatizes to one of these, so no change to the registry is needed — only the lemma has to be right before the gate. wot/wit is a lexical verb (VERB), not AUX.

UD English EWT 2.18 has one auxiliary art in its training set, annotated be|AUX|…|Person=2|Style=Arch|… (en_ewt-ud-train.conllu); the convention above follows it.

Feature consistency in the db (2026-10-17)

5.5 What the small model does

SLM now (2026-10-17), UPOS of the archaic auxiliaries in the 6 105 Torah sentences:

wilt is right because it is in EWT as the verb («wilt» = droop) and the modal reading follows from the context thou wilt + verb; everything else is a noun or an adjective to the model. 1 129 of the 6 013 aligned Torah sentences have a mis-tagged archaic auxiliary (SLM now). The lemma is now right thanks to the ONLY_VERBS list (shall, have, do whatever the tag), and since 2026-10-17 21:11 annot repairs the lemma of an AUX whose lemma is still the form (annot/src/main.rs, function drafts_from). After both fixes the refusals at the gate dropped from 28 % of the drafts (round 5: 1 050 of 3 791) to 0.4 % (round 9: 1 of 244) — chapter 12.

The remaining problem is that a noun shalt does not become an AUX with the right features by any deterministic step; the graph parser re-tags it and copies the features of the most frequent MD token. A closed-class table in front of the tagger (chapter 13, T1) removes the problem at its source.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/en-x-emode/05-auxiliaries/