Yezur Wiki:Ambiguous glosses

A working list of Dictionary entries whose English glosses do not distinguish them. It is the counterpart to a ruling the operator made on 2026-08-21 over calf and bone: where a language has two words and English has one, the one-word gloss inherited from the export hides the distinction rather than recording it, and the entries read as free variants when they are not.

Each row names the language, the competing lemmas, what their definitions currently say, and a suggested split. The suggestions are suggestions — the operator rules, and the ruling is what gets written.

How the list was made

Every row of the Lexemes table was grouped by language, part of speech and gloss; any group holding more than one lemma is a collision. Of 3,752 lexemes, 44 groups collided. Inflected-form pages, which carry no gloss, were excluded.

The collisions fall into three kinds, handled separately below: glosses where the definitions already draw the distinction and only the gloss field was stale; lemmas that genuinely need a ruling; and one case that is not an ambiguity at all but a mistake in an etymology.

Settled without a ruling

In these the entry already said the right thing and only the one-word gloss= field — which is what the collision check reads — was stale. Corrected 2026-08-21.

Language	Lemma	Was	Now	Because
Besoh	o'	fingernail	claw	The definition already reads "claw, talon (of beasts) — the human nail is kio'".
Besoh	go	ten	whole, full	The numeral sense is already marked archaic on the entry; it has passed to göj.
Besoh	bë	right	right (hand)	bë is the right hand and strong side; mon is right as in correct.
Besoh	mon	right	right (correct)	As above.
Proto-North-Andusian	*gira	play	game	From Proto-Andusian gida ("dance"), and glossed "contest, game" on the entry. The plain verb is gidz.

Dictionary:mum was also repaired in passing: its two Proto-South-Andusian etymologies had their part-of-speech headings at the same level as the etymologies, so neither nested under its own.

Part one: ruled and applied

The operator ruled every row on 2026-08-21 and the rulings are carried into the entries; the Notes column holds the ruling itself.

Herta-Olaic: one etymon, two reflexes

Thirteen Gaillean lemmas descend in pairs from a single Proto-Herta-Olaic word and reach Middle Airananwe still paired, with the same gloss on both. Two reflexes of one etymon is a normal thing for a language to have — but the pair almost always divides the sense-field between them, and here nothing records how.

#	Gloss	Proto	Gaillean	Middle Airananwe	Suggested split	Verdict	Notes
1	stone	*o:jig͡ɣ	oiat / oisie	oyte / oysye	Stone as a material against a stone as an object — the split already ruled for bone.	applied	oisie: piece of stone, oiat: material
2	sea	*kɸaɲɛx	piecia / qangha	piecie / qangyaa	The sea as an expanse against a named body of water; or inshore water against open water.	applied	synonyms
3	child	*ᵐbilwə	mbila / theste	mbile / these	A young person against someone's offspring, of any age.	applied	mbila: a child theste: offspring
4	burn	*rkaʒtə	awarei / reghazhtei	ooeree / regyasytee	Intransitive (to be alight) against transitive (to set alight).	applied	awarei: be on fire reghazhtei: transitive
5	live	*ruo:ʃt	rocettei / videi	rocetee / vindestee	To be alive against to dwell, to reside somewhere.	applied	rocettei: dwell videi: be alive
6	they	*aɪʃxe	aecei / icei	aycee / ecee	Animate against inanimate; or, given the pair below, the same distinction the singular makes.	applied	icei is female, aecei is male/generic.
7	he/she/it/etc	*xmɑtʲe:	athe / hima	—	As above. The gloss "he/she/it/etc" is itself worth replacing with whatever the ruling settles.	applied	hima is female, athe is male/generic.
8	you	ixaβ (pl.) / nwijɔ (sg.)	ihav / nio	hyaave / nie	Already answered upstream: the proto distinguishes plural from singular and Ancient Hertic labels its pair ιαπα (plural) / νου (singular). Gaillean and Middle need the same labels. See the error below.	applied	yes, clearly ihav and hyaave are plural, and nio/nie are singular.
9	in	*ho:tβ	hoth / ou	hoth / uue	Static location (in) against motion (into).	applied	hoth: inside of, contained by ou: more generically "at a location" -> in England.
10	at	*uztk	sou / uzh	—	Place against time; or a general locative against a precise one.	applied	sou: time uzh: place

Herta-Olaic: unrelated pairs

These three have no common etymon, so they are simply two words the export gave one gloss.

#	Gloss	Gaillean	Middle Airananwe	Suggested split	Verdict	Notes
11	by	ith / taw	ith / itaa	Agent ("written by her") against proximity ("by the door"). English merges these; most languages do not.	applied	taw: nearby / ith: action done by
12	for	bell / wiu	bel / iyu	Benefactive ("for you") against purpose or duration ("for three days").	applied	bell: purpose / iyu: benefactive
13	step-mother	siquiba / velna-hwith	siqie / velan-huith	velna-hwith is transparently a compound and siquiba is not, which usually means one is the plain word and the other formal, legal or distancing.	applied	siquiba is a more colloquial term, also meaning "adoptive mother" or "caretaker" or effectively "baby-sitter". velan-huith is the formal term.

Andusian

#	Gloss	Languages	Lemmas	Suggested split	Verdict	Notes
14	he, she, it	Proto-Andusian and all three branch protos, and Besoh	dʰɛ / o, and their reflexes	Two third-person pronouns run the length of the family with no distinction recorded anywhere. Proximal against distal, or animate against inanimate, would both be ordinary. Whatever is ruled propagates to eight entries.	applied	dhe: female o: male
15	hail	Proto-South-Andusian, Besoh	mum / rɔts, mium / roc	mum comes from muᵐb ("snow"), rɔts* from a word already meaning hail and ice. Soft hail or graupel against the hard hailstone would follow the etymologies.	applied	synonyms

Besoh: superseded senses left unmarked

Besoh repeatedly coined a new word to escape a homophone — and in one case, go / göj, the older entry says so plainly: "archaic: ten — the numeral sense has passed to göj." Four more pairs did the same thing without leaving the note, so both lemmas still claim the sense.

One ruling covers all four: whether the older word keeps the sense alongside the new one, or is marked superseded the way go is.

Human note: No, if there is a superseded form we don't keep the word on the old. This isn't divergent, it's just a differing path. If anything, we also don't list "lost final consonant restored" or "escaping homophones"; those are bordering on out-of-character comments.

#	Sense	Older lemma	Replacement	What the replacement's etymology says	Verdict
16	sharp	he (etymology 3)	heĥ	"the lost final consonant restored archaically"	applied
17	stone	ho (etymology 2)	hoĥ	"the lost final consonant restored archaically"	applied
18	spear	gö (etymology 3)	gön	"extended with -n to escape its homophones"	applied
19	new	ba' (etymology 2)	beza	"Origin uncertain" — so this pair may instead be a real split (ba' young, raw against beza new, fresh).	applied

An error, not an ambiguity

Gaillean ihav and its Middle Airananwe reflex hyaave both give their etymon as *nwijɔ — the singular — which is also the etymon of nio / nie. The plural *ixaβ has no Gaillean descendant at all, though Ancient Hertic ιαπα continues it and matches ihav in shape.

Reading ihav < *ixaβ fixes the etymology, gives the plural a reflex, and settles row 8 at the same time. It needs confirming rather than ruling, but it is a content change either way.

Human note: Confirmed - ihav and hyaave are from ixaB.

Part two: where English is the ambiguity

Filed 2026-08-21, after the operator's rulings above. The first list found places where one language had two words and English one gloss. This one finds the other half of the same problem: glosses that are a single English word which English itself splits. The operator's example is hair — the hair on a person's head, or one hair. Nothing in the corpus says which, so the gloss will eventually be translated wrong in both directions.

Every row of Lexemes was grouped by gloss; 258 single-word glosses are carried by four or more languages. The rows below are the ones where the English word does real damage, not the whole 258 — the rest are single-sense words (tooth, snake, five) that need nothing.

Fixed without a ruling

Housekeeping, corrected 2026-08-21 — these were inconsistencies in the gloss field rather than questions.

Was	Now	Affected	Why
`i` (Herta-Olaic) against `I` (Andusian)	`I`	4 entries	The first-person pronoun. Lowercase reads as the name of the letter.
`left` / `left(-hand)`	`left (hand)`	9 entries	The two halves of one family were glossed differently, and bare left can be read as a direction of travel or as "left over".
`right` / `right(-hand)`	`right (hand)`	9 entries	As above.
`right` (the Andusian moᵑg family)	`correct`	4 entries	These mean right as in correct, and now share the gloss the Herta-Olaic and Njeshan words for that sense already used.
`right` (*moᵑgets)	`the right, entitlement`	1 entry	It is the noun — "the right / law" — not the adjective.

Two different words, spelled alike

The calf case: English has two unrelated words and the corpus gave them one gloss. If ruled, each takes an English entry with two etymology headings and a table each, the way Dictionary:calf now does.

#	Gloss	Langs	The two English words	What the corpus has	Verdict
20	fly	12	The insect, and the verb.	Gaillean has both — connu (noun) and cehirei (verb) — each defined only as "fly". Every other language carries one or the other under the same gloss.	pending
21	stick	9	A stick of wood, and to stick or adhere.	All nine are the noun, but nothing says so.	pending
22	right	11	The hand or side, an entitlement, and correct.	Now split three ways in the gloss field (above), but the English entry, when it comes, needs the etymology headings.	pending
23	drink	5	To drink, and a drink.		pending
24	dream	4	To dream, and a dream.		pending
25	play	5	A game, music, and what children do. Most languages divide these.	Proto-North-Andusian already has two words, now play and game.	pending

Mass against count — the hair case

Here English uses one form for the stuff and for a piece of it. A language need not, and several in the corpus already do not.

#	Gloss	Langs	The split	What the corpus has	Verdict
26	hair	5	The hair on a head, against one hair.	Proto-Andusian mamgob says "hair (of the head)" and so is the mass; Proto-Herta-Olaic eɑkʲiəθ and its three reflexes say only "hair".	pending
27	stone	13	The material, a stone, a hailstone, a tool-stone, a grindstone.	Gaillean and Middle Airananwe are now split by ruling 1. The other nine are unqualified, and their definitions fan out across all five readings.	pending
28	bone	10	The substance, against a bone.	Split for Gaillean and Middle Airananwe (ettuz / eunduz); the rest unqualified.	pending
29	wood	7	The material, against a wood or grove.	Andusian sim is "wood, timber; firewood" — the material. English wood* also means a small forest, and the corpus has a separate forest gloss, so the two could collide.	pending
30	fruit, grass, rope, sand, ash, dust, salt, fat, meat, skin, snow, ice	4–10 each	Same shape throughout: the stuff, against one piece of it.	Unqualified. One ruling could set the house convention for all of them.	pending

Where the word covers a different extent

#	Gloss	Langs	The question	What the corpus has	Verdict
31	hand	6	Does the word stop at the wrist?	Proto-Andusian *ᵐbu is explicitly "the hand and forearm"; the four Herta-Olaic words say only "hand". If they mean the same thing, the gloss lies; if not, the English box is listing two different body parts.	pending
32	earth / ground	11 + 5	English ground translates both.	Andusian tse is "earth, soil; the ground as substance" and lem is "ground, floor; the low place, bottom". A translator reaching for "the ground" has no way to choose.	pending
33	tree / wood	9 + 7	Adjacent glosses on near-identical forms.	Andusian ciᵐb "tree" and sim "wood" fall together in the daughters — Proto-South-Andusian has ʃim "tree" beside sim "wood", Besoh śim beside sim. Whether that near-homophony is intended is worth a look.	pending
34	breast	10	The chest, against a woman's breast.	Unqualified.	pending
35	head	10	The body part, a leader, and the top of a thing.	Unqualified.	pending
36	ear	4	The organ, against an ear of grain.	Unqualified.	pending

Adjectives running on two axes

Each of these is one English word doing two jobs. Languages very often split them, and the corpus has no way to record which job a given word does.

#	Gloss	Langs	The two axes	Verdict
37	big	10	Size, against importance.	pending
38	long	9	Extent in space, against duration in time.	pending
39	short	5	Length, height and duration — three, not two.	pending
40	old	8	Aged, of a living thing, against not-new, of an object. Many languages have separate words.	pending
41	new	11	Lately made, against not met with before.	pending
42	dull	10	Blunt, not bright, and uninteresting. The Swadesh sense is blunt; nothing says so.	pending
43	thick / thin	5 each	Of a board (depth), of a rope (girth), of a person, of a liquid.	pending
44	heavy	5	Weight, against hard to bear.	pending
45	full	5	Filled, against complete. Besoh go is already "whole, full; the complete count".	pending

Distinctions English lacks, which the corpus may want

Not ambiguities in the gloss so much as decisions never taken. Listed because the answer changes what a translations box may contain.

#	Gloss	Langs	The distinction	Verdict
46	we	10	Inclusive against exclusive — "we, including you" against "we, not you". Very common; none of the ten is marked.	pending
47	know	10	To know a fact, against to know a person. Split in a great many languages.	pending
48	say	10	To say something, to tell someone, to speak a language. The corpus has say alone.	pending
49	smell	9	To give off an odour, against to perceive one. One gloss covers both.	pending
50	come / go	5 each	Whether motion is reckoned toward the speaker. The Andusian go family is now ruled as depart; come is unexamined.	pending