Nychthemeron

Tulkata Names

The sound system, the gazetteer, and a wolf nobody will name

Forms marked † are the author’s own and are fixed. Everything else is a proposal, drawn from seed 6001 against the phonotactics below; changing the seed replaces the whole proposed corpus. This page is generated — edit tools/build_tulkata.py and rerun.

Tulkata is written in the Latin letter with diacritics and so wants no conversion table. What it wants is an account of its sounds, because the attested material turns out to be remarkably consistent, and the consistency is what makes more of it possible.

I. The sound system

Derived from 38 words, 14 place-names and 4 personal names — some seventy forms.

There are no voiced stops. Not one b, d or g occurs anywhere in the attested material, and the second tranche of words confirmed it rather than contradicting it. The language has p t k and nothing behind them. This is the single most audible fact about Tulkata and the most useful constraint there is: it is what makes the tongue sound unlike every human speech in the setting, all of which have the voiced series.

Sounds
Stopsp · t · kvoiceless only
Affricatetzḥeu·tzun — read as a unit /ts/; see below
Fricativess · š · h · ḥ · łtwo h’s, and both common — see §IV
Nasalsm · n · ññ is frequent, and geminates (Pǫññilña)
Liquidsl · rl carries the language; it is the commonest sound in it
Glidesw · yrare, and never beside a like vowel
Vowels
Plaina · e · i · o · ua, u and i much the commonest
Nasalã · ǫstill only these two across the whole corpus
Longō · īmisrō, kulōn, milō · saiḥīl
Diphthongsau · ai · euvery frequent; weutkai has two

On tz. It occurs once, in ḥeutzun bog. Read as t + z it would be the only voiced consonant in the language and would spoil an otherwise perfect pattern; read as a unit /ts/ — which is what the spelling ⟨tz⟩ does in German and Polish — the obstruent system is wholly voiceless and the character of the tongue is intact. The second reading is assumed here. One more word with ⟨z⟩ in it would settle the matter either way.

A correction, and the lesson in it

An earlier draft of this page, written off the first thirty-six forms, declared five rules that the next sixteen words broke. They are set out because the correction is worth more than the rules were:

The rule I drewThe word that broke it
nothing begins with s-saiḥīl eagles- is rare, not absent
length only on ōsaiḥīlī exists
ḥ never after a consonantaumḥakai ravenmḥ is good Tulkata
rr impossibleNaiširrolan attested geminate after all
word-final codas only -m -n -lOtuñkǫh-h closes a word too

In a corpus of a few dozen forms, a gap is not a rule. The constraints are now drawn only where the evidence is positive, and the generator’s list of forbidden sequences has gone from thirty-odd to a handful. The one rule that survived — no voiced stops — is the stronger for having survived, because it has now been tested against twice the material and holds without exception.

II. Phonotactics

Syllable: (C)V(V)(C). No word in the corpus begins with a cluster, and none ends with one. Clusters arise only where a coda meets the next onset — Ãs·tã·him, ih·mu·sul, mah·lã, os·ñaum, kǫm·nu, łǫr·hōr, ul·sim, hil·na.

What the whole corpus now supports, stated as tendencies where that is all they are:

The evidence
Word-final codas run heavily to -m, -n, -lfourteen of the fifteen consonant-final forms, from ulsim and osñaum to Naiširrol and Nauñisutnil-h once, in Otuñkǫh; nothing ends in -s, -r or -ñ
s- is word-initial only oncesaiḥīl, against some seventy forms; s and š are common medially (Aši, ñuhhašã, misrō, Naiširrol)š- initially is still unattested
ł stands word-initially and nowhere elsełǫrhōr, its one attestationa marginal phoneme; still wants evidence
Attested geminates: hh · ññ · rrauhhi, ñuhhašã · Pǫññilña · Naiširrolthe stops never double
Long vowels are never word-initialmisrō, kulōn, milō, saiḥīl — all medial or finalheld as a rule, being a positive pattern

III. The gazetteer as it stands †

Four regions, each with its own country and its own character.

The second tranche of words parsed four names that had been opaque, and each one landed in the region where it belonged — which is the best evidence there is that the suffix analysis in §VI is right.

NameNow parses asAnd it stands in
Pǫññilñapǫññe ford + -il + -ñaÃstãhim, the swamp
Euḥãmulḥãmu island + -lÃstãhim — an island in a marsh is worth naming
LeutinoLeu- + tino peakYopuli, the mountains
Aiḥãšoaiḥã fox + -šoMaisweutkai, the forest

Not one of those glosses was available this morning, and not one had to be forced.

Ãstãhim † — the north-west

the swamp country. Does not parse, and should not be made to — see §VI.

Settlement
Aši †the chief town
Pǫññilña †pǫññe ford + -il + -ña — the ford-place
Euḥãmul †ḥãmu island + -l — the island-place, in a swamp, where an island is worth naming

Hauswawi † — the centre

the farming country. hau- is productive but unglossed — see §IV.

Settlement
Haukañan †the major town; hau- + kañan stone
Hutorma †a smaller town
Mauraiwauḥil †a smaller town
Paisayai †a smaller town
Ḥauseuye †a smaller town. Ḥ-, not H- — see §IV

Yopuli † — the south-east

yo- + puli mountain — the mountainous country.

Settlement
Tiyiḥsisil †a settlement; reduplicating si-sil — see §VI
Leutino †Leu- + tino peak — in the mountain region, as it should be

Maisweutkai † — the north

the wolf-wood: mais- + weutkai forest. See §V — mais- is not the living word for wolf.

Settlement
Aiḥãšo †aiḥã fox + -šo — the fox-place

IV. The trap: h is not ḥ

Both are common, both are frequent word-initially, and they are different sounds. The gazetteer sets the trap itself:

Hauswawi · Haukañan · Hutormaplain hthe farming country and its towns
Ḥauseuyeḥa different word, in the same region

Hauswawi, Haukañan and Ḥauseuye sit within a few leagues of one another and the first two share an element the third does not. A foreign scribe who hears one h will write all three alike and produce three villages where there are two families of name — which is exactly the error to expect from an imperial surveyor, and exactly the error the Great Settlement’s registers would be full of. There is a table in that, and a lawsuit.

And hau- wants a gloss. It heads the farming region and its chief town (Hau·kañan = hau- + kañan stone), so it is productive and it is not in the word-list. Field, tilled land or cleared ground would suit the country and would make Haukañan the boundary-stone in the field — but it is the author’s to settle, and so is the yo- of Yopuli.

V. The wolf in the wood

Maisweutkai is the wolf-wood. Its second element is weutkai forest, plainly. Its first is mais-.

But the Tulkata word for wolf is mahlã.

Those two cannot be made into one another by any sound-change the rest of the corpus supports, which leaves one explanation, and it is the better one:

mais- is the older word for wolf, and it survives in this name and nowhere else. mahlã replaced it — and the replacement of a predator’s name is the best-attested kind of lexical substitution there is. Our own bear is the brown one, and Slavonic медведь is the honey-eater, because the beast’s true name was not to be said aloud.

Which fits the Tulkata exactly as they already stand. They are of the Children of the Wilds, they hold the world finished, and their gods are wild ones. A people like that do not say the wolf’s name.

And so:

VI. How names are made

Region names are head-final: modifier + geographic generic. Two of the four parse outright — Yo·puli (mountain) and Mais·weutkai (forest) — and that is the pattern to coin on. The generics available are puli · weutkai · laumu · hilna · milir · auhhi · lailau · rula.

Ãstãhim and Hauswawi do not parse, and should not be made to. A map on which every name resolves is a map with no history in it; the swamp country in the north-west is the likeliest place for a name older than the language, on the same principle that keeps Obana in Ascania and the Ukkul layer in Pavasia.

Settlement names run to four patterns, in roughly the proportions the attested names show:

PatternAttested
root + place suffixEuḥãmul · Tiyiḥsisil · Hutorma · Leutinothe commonest
modifier + geographic headHaukañanas the regions are made
reduplicationPaisayai · Hauswawisee below
a bare rootAšichief towns, chiefly

The place suffixes

-Vl is far the commonest, and it appears on a common noun as well — ihmu·sul, mountain pass — which suggests it was a live formative and not only a fossil in names. The glosses are proposals; the forms are drawn from the gazetteer.

SuffixProposed senseIn
-ulplace ofEuḥãmul, ihmusul
-ilplace of (after a front vowel)Mauraiwauḥil
-silplace of (extended)Tiyiḥsisil
-ñaamong the, the company ofPǫññilña
-mathe dwelling, the steadingHutorma
-othe little, the lesserLeutino, Aiḥãšo
-ithe (bare definite)Aši, Hauswawi
-aithe folk ofPaisayai

Reduplication

The reduplicant copies the base’s onset and takes -ai- before an open syllable and -i- before a closed one. Two attestations give the rule and agree on it:

ReduplicantBase
lai·lau plainslai-lauopen base → -ai-
Tiyiḥ·si·silsi-silclosed base → -i-

Hauswawi and Paisayai carry the same shape. If lailau plains is a reduplication of an unrecorded lau, then that root is a further fossil and wants a sense — the open land, presumably, which would make plains literally the spreading-out.

VII. Uzda, and the one letter that settles it

The archive already holds a troll name of exactly this shape. Svelekebá, the Tulkata kingdom the Great Settlement is surveying, is glossed in the archive’s own terms as Uzda svele, bear, and kebá, forest — the bear-wood.

Set it beside the new material and two things are plain at once.

UzdaTulkata
bearsveleulsim
forestkebáweutkai
the compoundSvele·kebá — the bear-woodMais·weutkai — the wolf-wood

The syntax agrees exactly and the vocabulary agrees not at all. Modifier first, geographic head last, in both — which is the signature of tongues related in their grammar and divergent in their words, and not of one language with two registers.

And then there is the b. Kebá has a voiced stop. Tulkata has no voiced stops at all — not one b, d or g in thirty-six forms. So whatever else is true, kebá cannot be a Tulkata word, and the natural reading is a devoicing: the ancestral b d g fell together with p t k in Tulkata and did not in Uzda. One letter in one gloss, and it dates the relation.

Which leaves the author three readings, and the b survives all of them — Uzda as the elder stage, Uzda as a sister tongue, or Uzda as no relation at all. The compendium’s standing assumption, that the Uzda tongues simply are Tulkata speech, is the one reading the evidence will not bear.

It also disposes of the provisional-name problem without discarding the etymology. Svelekebá need not be renamed: let it be the Uzda name and give the Tulkata name beside it — Ulsimweutkai, or a trimmer Ulsiweutkai — so that the Empire’s chancery and the kingdom’s own people call it different things. That is the Baran/Weissenburg device, and the Nieuwdorp/Neudorf device, arriving for the third time in a different family; and a surveyor under the Great Settlement who records both will enter one forest twice.

And it answers §V’s question before it was asked. I wondered there whether an older bear-word lay fossilised in some place-name not yet written. It is written already, and it is svele — the wolf-wood and the bear-wood, each keeping a beast’s name that its own speakers have otherwise stopped saying.

VIII. The lexicon

38 attested words †, and 30 proposals drawn to the phonotactics above. Every proposal stands at least two edits clear of every attested form — words, places and persons alike — so nothing in the right-hand column can be mistaken for something in the left.

Six senses were struck from the proposed list when the author supplied them himself — island, bog, raven, birch, oak and eagle. Six others took their place.

Marsh and water

Attested †Proposed
omlu † — bayãseḥo — mist, fog
ḥeutzun † — bogiñnã — reed
leupin † — fishkuma — peat
ḥãmu † — islanduhhu — eel
hilna † — lakemeshai — heron
auhhi † — marshmirpu — otter
leurim † — meadowyaiño — alder
ñuhhašã † — mossenai — fen
lamãnu † — mushroom
laumu † — river
łǫrhōr † — willow

Field and pasture

Attested †Proposed
lima † — daleeuti — barley
milō † — deerłelōhǫ — ox
kǫmnu † — fordleupa — furrow
pǫññe † — ford, crossing (beside kǫmnu)uhonpa — hearth
osñaum † — grasshoslǫmwa — mill
leurim † — meadowḥãyã — byre, cow-house
lailau † — plainshuyã — boundary
laumu † — riverpōlaimne — frost
pasan † — rock
kañan † — stone
rula † — valley

Rock and height

Attested †Proposed
ulsim † — bearhauni — snow
kulōn † — cliffmãḥta — ice
lima † — dalemamō — scree
saiḥīl † — eagleḥemimo — goat
ḥušau † — hillholalil — cave
puli † — mountainñeurḥǫl — spring, source
ihmusul † — mountain passlutō — thunder
tino † — peakletzan — ridge
misrō † — pine
pasan † — rock
kañan † — stone
rula † — valley

Wood and beast

Attested †Proposed
ulsim † — bearnamau — bark
ilsi † — birchomlin — boar
luñlihi † — cat, lynxpañḥo — ash (tree)
weutkai † — forestyãmal — honey
aiḥã † — foxãslauhi — antler
ñuhhašã † — mossermu — clearing
lamãnu † — mushroom
ḥansau † — oak
misrō † — pine
aumḥakai † — raven
ñaḥsa † — spruce
maurli † — thicket, small wood
mahlã † — wolf

A word may stand in more than one field — laumu, kañan and rula are common to several, which is why they repeat.

IX. The naming of persons

Four names, and they are enough to show the shape of the thing.

Name †
Weyaumḥakai †Wey- + aumḥakai raven — a man named for a beast
Naiširrol †ends in -ol; the corpus’s only rr
Nauñisutnil †ends in -il, as several towns do
Otuñkǫh †vowel-initial, and the corpus’s only final -h

Personal names run longer than place-names — four syllables and upward against the gazetteer’s two and three — and they close in -ol, -il, -ul or, once, -kǫh.

And a man may be named for a beast. Wey·aumḥakai carries aumḥakai, the raven, entire. Which raises the question §V has already answered: nobody is named for the wolf. If mahlã is a name not said plainly, it cannot stand in a child’s name either — so the list of beasts a man may be called after is the list of beasts that are not under avoidance, and the generator below is restricted accordingly: aumḥakai · ulsim · milō · luñlihi · leupin · aiḥã · saiḥīl. A Tulkata called for the wolf would be a remarkable man, and the name would mean something had happened.

Twenty-four proposals, on the three patterns the attested names show — an element plus a beast, a long root plus a personal suffix, and a bare three-syllable root. No beast is carried by more than two men.

Mǫleḥol · Ḥihaiḥã · Sãmpulsim · Ǫnsirlōnrol · Ḥeulaiḥã · Ǫhtzanul · Kaulluñlihi · Masemeukǫh · Rihkihtil · Loneunirol · Kãlḥolmorol · Łãsmōrko · Kelawil · Pollisaiḥīl · Laulaumḥakai · Walãmno · Sãpimãrol · Eumaunul · Riñḥašil · Mailmilō · Paululsim · Kunawīrol · Eunḥurol · Kaummilō

What is wholly absent is any sign of how they are borne: no patronymic, no clan element, no title. The Emperor still has no name, and now there is nothing stopping him having one.

X. Proposed settlements

Seven apiece, drawn with each region’s own vocabulary so that the names sound local: marsh words in Ãstãhim, rock and height in Yopuli, and so forward. No geographic head is used twice in one region.

Ãstãhim — the north-west, marsh and water

Airlima · Kišeleurim · Raillaumu · Kaummema · Yaññaleupin · Lailauršǫ · Paiposu

Hauswawi — the centre, field and pasture

Ḥeukauhhi · Ãmhuma · Nama · Naunleurim · Aurhaña · Auleril · Rǫli

Yopuli — the south-east, rock and height

Rumessiña · Leluḥlai · Mono · Kaiši · Laulhilna · Ḥitzo · Lumentino

Maisweutkai — the north, wood and beast

Laisai · Ñillinul · Ñiñakom · Arosḥauma · Ankinnil · Raiḥãnasil · Ñahaisil

Those that parse do so against the lexicon above; those that do not are the older stratum, and are worth keeping unparsed in the proportion they come out.

What is still wanting

See also

Tulkata · The Two Powers · Races · Geography · Pavasian Names · Azadean Names · Ascanian Names