who, whom, whose: case and the lemma who (English)

Gist. After 2.18 the English PRON.md closed two TODOs. Who pronouns got the Case feature: who — Nom or Acc by context, whom — Acc with lemma who, dependent whose — Case=Gen|Poss=Yes, standalone whose — only Poss=Yes. Likewise whoever, whomever (lemma whoever) and the new form whosever.

Level. English, _en/pos/PRON.md. A guideline, not the validator: the feature registry for en did not change.

Before (2.18). 2.18:docs/_en/pos/PRON.md:48 — «PronType is the only feature except where shown below». TODO at :66 (whom(ever): case and lemma who) and :68 (whose: Poss=Yes, possibly Case=Gen). Details — The pronouns who, whom, whose in 2.18: TODO in the guideline.

Now (snapshot 24.09.2026).

docs/_en/pos/PRON.md:48: «Case is marked only for who and variants»;

:50–54 — a table:

	`Case=Nom`	`Case=Acc`	dependent possessive: `Case=Gen`, `Poss=Yes`	independent possessive: `Poss=Yes`
`PronType=Int` or `Rel`	who	whom [who], who	whose	whose
`PronType=Int` or `Rel`	whoever, whosoever	whomever [whoever], whoever	whosever	—
XPOS	WP	WP	WP$	WP$

:56: «who and whoever may bear either value of Case, depending on context. The variant whosoever receives Style=Form»;
the TODOs about whom and whose were removed; TODOs remain about whatsoever, whatever as DET and exclamative what (:68–72).

Evidence. diff 2.18:docs/_en/pos/PRON.md ↔ docs/_en/pos/PRON.md: lines 48–69. Context — #517 «Lemmas of English personal pronouns» (closed 19.11.2025, milestone v2.18): «I am not sure why whom, whomever … aren't normalized to nominative». There is no separate issue about the change itself in the dump. Treebank branches are not in the snapshot, so whether EWT dev has already reworked this is unknown.

What it means for English annotation.

lemma who for all forms, as in GUM, GENTLE, LinES and ParTUT. EWT 2.18 has lemma whom (23);
case in FEATS for 447 EWT tokens: who 422 (one of them a typo with lemma whoever), whom 23, whoever 1, whomever 1;
on whose (13) — Case=Gen: all 13 in EWT are nmod:poss, i.e. dependent;
Case for who is taken from the syntactic function. This agrees with the 2.19 clarification on features from context (Features from context and Exponence in MISC (universal, 2.19)). The form whom gives Acc even in subject position, because the form weighs more.

Registry. No new pairs needed: feat PRON Case=Nom|Acc|Gen and feat PRON Poss=Yes are already in en/data/ud-registry-en.tsv. Only Style=Form for whosoever is missing, but that was so in 2.18 too, and EWT has no such form.

Sources. docs/_en/pos/PRON.md, 2.18:docs/_en/pos/PRON.md; #517, #1009 (independent possessives); seed Interrogative and relative pronouns: who, whom, whose, which, what, that (en.morph.whom-acc is already written according to this guideline).

EWT 2.18 does not conform to these rules, and that is the negative control: the error rules give 24 (whom) + 423 (who without Case) + 13 (whose) violations, the warn rules 7 + 6 + 396. After the mova → ud-next converter (Converter mova ⇄ ud-next) — 0, checked with the engine on 26.09.

rule: udnext.whom-who-acc
what: whom/whomever — lemma who/whoever and Case=Acc
match: w[upos=PRON, form=whom|whomever]
require: w[lemma=who|whoever, feats.Case=Acc]
severity: error
source: UD docs/_en/pos/PRON.md:52–53 (snapshot 24.09.2026)

rule: udnext.who-case
what: who/whoever always has Case — Nom or Acc by context
match: w[upos=PRON, form=who|whoever]
require: w[feats.Case=Nom|Acc]
severity: error
source: UD docs/_en/pos/PRON.md:48, :56 (snapshot 24.09.2026)

rule: udnext.who-object-acc
what: who/whoever as an object — Case=Acc
match: w[upos=PRON, form=who|whoever, rel=obj|iobj]
require: w[feats.Case=Acc]
severity: warn
source: UD docs/_en/pos/PRON.md:52, :56 (snapshot 24.09.2026)

rule: udnext.who-oblique-acc
what: who/whoever as a prepositional object (who … to, who … with) — Case=Acc; the head of a free relative is left alone
match: w[upos=PRON, form=who|whoever, rel=obl|obl:agent|nmod]
require: w[feats.Case=Acc]
unless: exists r[rel=acl:relcl, head=w]
severity: warn
source: UD docs/_en/pos/PRON.md:52, :56 (snapshot 24.09.2026)

rule: udnext.who-subject-nom
what: who/whoever as a subject — Case=Nom
match: w[upos=PRON, form=who|whoever, rel~nsubj]
require: w[feats.Case=Nom]
severity: warn
source: UD docs/_en/pos/PRON.md:52, :56 (snapshot 24.09.2026)

rule: udnext.whose-dependent
what: whose with a noun — Case=Gen and Poss=Yes
match: w[upos=PRON, form=whose, rel=nmod:poss]
require: w[feats.Case=Gen, feats.Poss=Yes]
severity: error
source: UD docs/_en/pos/PRON.md:52 (snapshot 24.09.2026)

rule: udnext.whose-independent
what: standalone whose — Poss=Yes without Case
match: w[upos=PRON, form=whose, rel!=nmod:poss]
require: w[feats.Poss=Yes, !feats.Case]
severity: warn
source: UD docs/_en/pos/PRON.md:52 (snapshot 24.09.2026)

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/dialects/ud-next/seeds/en-wh-case/