bradrn’s scratchpad

[bradrn]

Now I've finished reading it. How long did it take you to make this? I never get that much done on grammar.

Well, you posted the challenge in May, and I posted the language in July, so… two months? But I didn’t work on it consistently.

This is not fixed stress location though...

It’s fixed given the underlying form.

Is this how it's usually done? I mean in natphonology.

Well, I don’t recall ever seeing it in a natlang. But hey, it’s an interesting system!

[Qwynegold]

Well, you posted the challenge in May, and I posted the language in July, so… two months? But I didn’t work on it consistently.

That's fast for that amount of work!

Well, I don’t recall ever seeing it in a natlang. But hey, it’s an interesting system!

Oh? Well that's well done for coming up with such a system on your own.

[bradrn]

Well, you posted the challenge in May, and I posted the language in July, so… two months? But I didn’t work on it consistently.

That's fast for that amount of work!

Really? I thought I was being slow. But thanks!

Well, I don’t recall ever seeing it in a natlang. But hey, it’s an interesting system!

Oh? Well that's well done for coming up with such a system on your own.

Actually, now that I think about it, I did see a comment about how phonological rules can distort stress assignment in some languages. But it didn’t go the the extremes I went with Wēchizaŋkəŋ.

[bradrn]

Interlude: some linguistic geography

While I’m still working on an outline of the Hlʉ̂ verb complex (which at this rate looks like it may never be finished), here’s a very rough map to put what I’ve been talking about into context:



For perspective, the enclosed sea should be about the same size as the Mediterranean; the island chain is comparable to Indonesia. In terms of climate, the bottommost strip of the map is meant to be rainforest, graduating to temperate at the very top. Practically everything here is subject to change, though.

As for the languages: starting from the north (yes, I know I forgot to put a compass rose on there), we have the Hlʉ̂ish languages, naturally including Hlʉ̂ itself. Its distribution as shown on the map actually is fairly implausible considering the size of the island chain; more sensible would be for the Hlʉ̂ to originate at the south end of the island chain and back-colonise the more northern islands. They maintain a thalassocracy: we can already see a few Hlʉ̂ colonies on the mainland. I suspect the Hlʉ̂ish Urheimat was on the east coast, whence they were displaced northwards.

Moving clockwise, the ‘Savanna’ family is currently rather vaguely defined — in fact, I have a horrible feeling that I’ve drawn it in a place with no savannas at all, so the name may have to be changed (hence the question mark on the map). It’ll be widely-distributed and large though, around the size of IE or Oto-Manguean. ‘Æŋ’ (again, the name is uncertain) may or may not be a highly divergent ‘Savanna’ language.

I should note that Proto-‘Savanna’ is actually attested in writing, admittedly somewhat fragmentarily. I didn’t bother naming it at the time, but that’s the language I translated The North Wind and the Sun into. I’m actually really unsatisfied with most of the key areas of its grammar, and plan to redo it substantially soon — this time, posting a proper grammar as I go. (It might even get posted before we see the Hlʉ̂ verb complex! Not that that’s hard; pigs might fly before we see the Hlʉ̂ verb complex.)

Finally, down the bottom left of the map is some sort of mountainous rainforesty area. Like most mountainous rainforesty areas, this is a region of considerable linguistic diversity. As noted on the map, Wēchizaŋkəŋ originates from this area, but I have no idea what the other languages there will look like. It would be nice if this region were to maintain this diversity into the modern era, but I suspect that’s implausible, what with it being as close to civilisation as it is.

As for more interesting topics, the only other thing I can say is that around the south there is a Sprachbund, centred on ‘Æŋ’, Wēchizaŋkəŋ and Hlʉ̂. (This also happens to be the most urbanised area on the map — or, perhaps, the only urbanised area on the map.) At the pictured time it’s fairly weak: the only obvious trait they share is phonological, since they all have experienced syncope and umlaut. Otherwise, the three languages are currently so different that it’s hard to know where this Sprachbund could end up. I’m thinking that possibly Hlʉ̂ could become more agglutinative and lose its converbs, Wēchizaŋkəŋ could fully monosyllabise its roots and get verb compounding, and ‘Æŋ’ could… um… stay the same, I guess, which would lead to three quite similar languages, now that I think about it. Randomly inventing some sample sentences for neo-Hlʉ̂ and post-Wēchizaŋkəŋ to compare to an ‘Æŋ’ sentence I currently have lying around, it would probably look something like this:

‘Æŋ’: sæssæŋ šɔ ælsæŋ muŋiwæyelle ‘wind-DEF and sun-DEF ???-do-argue-CONT’ [sorry, forgot how I glossed this one; I think ‘???’ were person prefixes]
neo-Hlʉ̂: seh hlʉ́ sîp ʔed hlʉ́ pomlɵlaye ‘wind CL go-SEQ sun CL try-talk⟨PL⟩-3’
post-Wēchizaŋkəŋ: ɣan nja want wimanaɣzəməɣna ‘wind sun COM 3PL.O-PFV-MED-speak-differ’

[keenir]

Interlude: some linguistic geography

While I’m still working on an outline of the Hlʉ̂ verb complex (which at this rate looks like it may never be finished),

sometimes the best works take a fair bit of time - as evidenced by your statement.

here’s a very rough map to put what I’ve been talking about into context:

thats a nice map.

As for the languages: starting from the north (yes, I know I forgot to put a compass rose on there), we have the Hlʉ̂ish languages, naturally including Hlʉ̂ itself. Its distribution as shown on the map actually is fairly implausible considering the size of the island chain;

I'm not sure I understand/follow...Java and Japan and England are all island nations that spread out to their neighboring islands and beyond (and Hawaii even discovered New Zealand, so distance isn't a problem)

Or did you mean that its implausible that the Hluish language family(s) cover that broad of an area? (for which, I point to PIE and Polynesian as evidence to the contrary)

‘Æŋ’ (again, the name is uncertain

maybe its just how the name was relayed to the Hlu? (like how o'api became okapi and thats what got passed on to Europeans curious about the "forest zebra" skins.

It would be nice if this region were to maintain this diversity into the modern era, but I suspect that’s implausible, what with it being as close to civilisation as it is.

Probably not terribly difficult...Indonesia and India haven't lost too many languages, despite having had cities for centuries. (I figured Brazil and Australia were probably not the best examples to cite)

‘Æŋ’: sæssæŋ šɔ ælsæŋ muŋiwæyelle ‘wind-DEF and sun-DEF ???-do-argue-CONT’ [sorry, forgot how I glossed this one; I think ‘???’ were person prefixes]
neo-Hlʉ̂: seh hlʉ́ sîp ʔed hlʉ́ pomlɵlaye ‘wind CL go-SEQ sun CL try-talk⟨PL⟩-3’
post-Wēchizaŋkəŋ: ɣan nja want wimanaɣzəməɣna ‘wind sun COM 3PL.O-PFV-MED-speak-differ’

quite nice.

[bradrn]

here’s a very rough map to put what I’ve been talking about into context:

thats a nice map.

No it isn’t.

As for the languages: starting from the north (yes, I know I forgot to put a compass rose on there), we have the Hlʉ̂ish languages, naturally including Hlʉ̂ itself. Its distribution as shown on the map actually is fairly implausible considering the size of the island chain;

I'm not sure I understand/follow...Java and Japan and England are all island nations that spread out to their neighboring islands and beyond (and Hawaii even discovered New Zealand, so distance isn't a problem)

Or did you mean that its implausible that the Hluish language family(s) cover that broad of an area? (for which, I point to PIE and Polynesian as evidence to the contrary)

I meant that it’s implausible for the whole island chain to be covered by a single language (Hlʉ̂).

‘Æŋ’ (again, the name is uncertain

maybe its just how the name was relayed to the Hlu?

No, it’s an endonym. Hlʉ̂ has neither /æ/ nor /ŋ/. (It would probably end up borrowed as something like En, if I stick with the name.)

It would be nice if this region were to maintain this diversity into the modern era, but I suspect that’s implausible, what with it being as close to civilisation as it is.

Probably not terribly difficult...Indonesia and India haven't lost too many languages, despite having had cities for centuries. (I figured Brazil and Australia were probably not the best examples to cite)

But India is almost completely covered by Indic and Dravidian, two very homogenous language families, and Indonesia is even more homogenous in this regard if you ignore its Papuan territories.

[keenir]

here’s a very rough map to put what I’ve been talking about into context:

thats a nice map.

No it isn’t.

well i like it and think its a nice map.

As for the languages: starting from the north (yes, I know I forgot to put a compass rose on there), we have the Hlʉ̂ish languages, naturally including Hlʉ̂ itself. Its distribution as shown on the map actually is fairly implausible considering the size of the island chain;

I'm not sure I understand/follow...Java and Japan and England are all island nations that spread out to their neighboring islands and beyond (and Hawaii even discovered New Zealand, so distance isn't a problem)

Or did you mean that its implausible that the Hluish language family(s) cover that broad of an area? (for which, I point to PIE and Polynesian as evidence to the contrary)

I meant that it’s implausible for the whole island chain to be covered by a single language (Hlʉ̂).

hmm...maybe Hlu is a koine, or - *has idea*

...Maybe an Amazonian-style system holds sway, of spouses and allies being preferred if they speak more than just the same language as the person considering the spouse/ally....and Hlu is simply the most the language there with the most speakers? (does that make sense?)

‘Æŋ’ (again, the name is uncertain

maybe its just how the name was relayed to the Hlu?

No, it’s an endonym. Hlʉ̂ has neither /æ/ nor /ŋ/. (It would probably end up borrowed as something like En, if I stick with the name.)

probably should have cited "naranj" and "orange" rather than "o'api" and "okapi"

...what the Hlu do when faced with that /ŋ/ that was relayed to them by their neighbors, is, yeah, another matter.

It would be nice if this region were to maintain this diversity into the modern era, but I suspect that’s implausible, what with it being as close to civilisation as it is.

Probably not terribly difficult...Indonesia and India haven't lost too many languages, despite having had cities for centuries. (I figured Brazil and Australia were probably not the best examples to cite)

But India is almost completely covered by Indic and Dravidian, two very homogenous language families, and Indonesia is even more homogenous in this regard if you ignore its Papuan territories.

My point was less about homogeny(sp), and more about languages not being run over. apologies for not being clear.

[bradrn]

hmm...maybe Hlu is a koine

No, it’s a very standard sort of language.

...Maybe an Amazonian-style system holds sway, of spouses and allies being preferred if they speak more than just the same language as the person considering the spouse/ally....and Hlu is simply the most the language there with the most speakers? (does that make sense?)

I don‘t think so. Honestly, I think I can get reasonably close to the situation depicted on the map just through normal historical processes.

Probably not terribly difficult...Indonesia and India haven't lost too many languages, despite having had cities for centuries. (I figured Brazil and Australia were probably not the best examples to cite)

But India is almost completely covered by Indic and Dravidian, two very homogenous language families, and Indonesia is even more homogenous in this regard if you ignore its Papuan territories.

My point was less about homogeny(sp), and more about languages not being run over. apologies for not being clear.

I should have been clearer also. I wanted not just a high density of languages, but high diversity as well:

[keenir]

I should have been clearer also. I wanted not just a high density of languages, but high diversity as well:

hmm...then maybe Australia (and Canada and the Congo?) might actually be better suggestions than I thought they'd be.

[bradrn]

Proto-‘Savanna’: Phonology and word classes

As briefly mentioned earlier, I have decided to completely redo this language. Its current version is serviceable — I even posted a translation of The North Wind and the Sun into it — but I’m decidedly unsatisfied with most parts of it. Probably the part I’m most unhappy about is its sadly underdeveloped coverb system, but there are problems throughout its phonology, romanisation, morphology, syntax and lexicon. By this stage it badly needs a full relexification and regrammaticalisation.

I’ve actually been meaning to do something like this for quite a while. The proximal catalyst here, however, was vegfarandi’s post on phrasal verbs in Duraic, which implements a closed verb system with coverbs a lot better than I did, and certainly does what I wanted to do better than I did. This felt decidedly unfair, especially given the amount of time I spent researching closed verb classes. How could I let this stand?

(Oh, and by the way, Proto-‘Savanna’ is a temporary name. As I said, they probably didn’t even live in the savanna.)

Phonology

Consonant inventory

Phonemic consonants are as follows:

	Labial	Alveolar	Lateral	Palatal	Velar	Glottal
Nasal	m ᵐb	n ⁿd			ŋ ᵑɡ
Voiceless stop	p pʰ	t tʰ			k kʰ	ʔ
Voiced stop	b	d
Affricate		t͡s t͡sʰ	t͡ɬ t͡ɬʰ
Fricative	f	s	ɬ		ɣ	h
Approximant			l	j	w

In slightly more detail:

• The nasals exist at bilabial, alveolar and velar PoAs. All voiced stops can be prenasalised; in fact, the voiced velar stop exists only in its prenasalised form /ᵑɡ/.
• There are four stop series: voiceless, aspirated, prenasalised (already mentioned) and voiced. Generally the aspiration is strong, and voiced stops are fully voiced, occasionally slightly implosive. The glottal stop exists only in one series, conventionally taken to be voiceless; only three series of velar stop exist, as there is no fully voiced velar stop */ɡ/.
• The four affricates behave similarly to the stops, but exist only in voiceless and aspirated series.
• The fricatives exist at labiodental, alveolar, lateral alveolar, velar and glottal PoAs. The velar fricative /ɣ/ is voiced, whereas all other fricatives are voiceless. /f h/ are noticeably rarer than the other fricatives in the lexicon, though at least /h/ occurs in a few particularly common words (e.g. /mah/ ‘do.PFV’).
• There are three approximants, /l j w/. Note that there are no rhotics. The velar fricative /ɣ/ can occasionally take on an approximant realisation [ɰ] too.

Vowel inventory

There are four short vowels, /a ə i u/, and two long vowels, /aː iː/. In ideophones, /uː/ may also occasionally be found. As usual for small vowel systems, the phonetic range of these vowels is subject to variation:

• /a/ may be any low vowel, though most commonly it is realised around [æ] or occasionally [ɐ] or [ɑ]
• /ə/ is usually simply the schwa [ə], but it may be backed to [ɤ̞] or fronted to [e̞] depending on the surrounding consonants; it may be any mid unrounded vowel
• /i/ is any high, unrounded, non-back vowel: usually it is realised as [ɪ], or sometimes [ɨ]
• /u/ is any back rounded vowel, almost always [ʊ]
• /aː/ is a non-high, unrounded, long vowel, usually centralised; commonly realised as [ɐː], but can be [ɜː] or even [əː] on occasion
• /iː/ is a high, front, long vowel, almost always [iː]
• /uː/ is rare, and always realised as [uː]

Phonotactics and stress

Syllables are at least CV, and may be CVC, where C represents any consonant and V any vowel. There appear to be no restrictions as to which consonants may occur in the onset or coda. However, all syllables have an onset; hence vowels never occur in hiatus.

Perhaps more significant are constraints on word composition. Content words are at least dimoraic: they are either disyllabic, or contain a long vowel or a coda (or both). Notably, the vast majority of content words are disyllabic, and follow a CVCV or CVCVC pattern. Only a handful of content words are monosyllabic. By contrast, there exist numerous monosyllabic grammatical words and particles, many of which are in fact only one mora. These may in fact be clitics, though wordhood is difficult to determine.

Romanisation

For now I’ve been using the following romanisation:

/m ᵐb d ⁿd ŋ ᵑɡ/ ⟨m mb n nd ŋ ng⟩
/p pʰ t tʰ k kʰ ʔ/ ⟨p ph t th k kh q⟩
/b d (ɡ)/ ⟨b d (ġ)⟩
/t͡s t͡sʰ t͡ɬ t͡ɬʰ/ ⟨ts tsh tl tlh⟩
/f s ɬ ɣ h/ ⟨f s lh g h⟩
/l j w/ ⟨l y w⟩
/a ə i u aː iː (uː)/ ⟨a e i u aa ii (uu)⟩

However, I can’t say I love this one: it’s ugly, and a bit difficult to read too. Lately I’ve been thinking about some alternatives, and this is what I’ve come up with so far:

Current	Sasat thaŋ mu tlaqufli qalit thaŋ waqli gilut fawetlli, naqeŋ khayiqene bey tshaal pawtiq, qaŋeth-inga siwitsi lamaq yusayetsi.
Americanist	Sasat thaŋ mu ƛaʔufli ʔalit thaŋ waʔli ɣilut fawəƛli, naʔəŋ khayiʔənə bəy chaꞏl pawtiʔ, ʔaŋəth-inga siwici lamaʔ yusayəci.
IPA-inspired	Sasat thaŋ mu tɬaʔufli ʔalit thaŋ waʔli ɣilut fawətɬli, naʔəŋ khajiʔənə bəj tshaːl pawtiʔ, ʔaŋəth-iŋga siwitsi lamaʔ jusajətsi.

And the same, with inferrable instances of /ʔ/ removed:

Current	Sasat thaŋ mu tlaufli alit thaŋ waqli gilut fawetlli, naeŋ khayiene bey tshaal pawtiq, aŋeth-inga siwitsi lamaq yusayetsi.
Americanist	Sasat thaŋ mu ƛaufli alit thaŋ waʔli ɣilut fawəƛli, naəŋ khayiənə bəy chaꞏl pawtiʔ, aŋəth-inga siwici lamaʔ yusayəci.
IPA-inspired	Sasat thaŋ mu tɬaufli alit thaŋ waʔli ɣilut fawətɬli, naəŋ khajiənə bəj tshaːl pawtiʔ, aŋəth-iŋga siwitsi lamaʔ jusajətsi.

Since the current orthography was decided by popular vote, I’ll do the same again. Which of these romanisations, or combinations thereof, does everyone most prefer? For now I’ll stick with my current one, but feel free to reply!

Word classes

A proper understanding of Proto-‘Savanna’ grammar is impossible without being aware of its unique word-class system. (Well, nearly unique; some natlangs come close.) Briefly summarising the major word classes:

• Nominals, which can head a noun phrase and act as a verbal argument; consisting of:
  • Common nouns, which refer generally to a type of object or concept;
  • Proper nouns, which refer to a specific person, place or object;
  • Adjectives, which can modify nouns and act as verbal adjuncts;
  • Numerals, which indicate the number of a corresponding noun; and
  • Personal pronouns and some demonstratives, which have deictic reference.
• Verbs, which head a verb phrase and can take aspectual markers.
• Coverbs, which combine with specific verbs to express new meanings.
• Adverbs, which modify verbs.
• Adjuncts, which are placed in a peripheral position and modify whole sentences.

Cross-cutting the above classes are demonstratives, which may be nouns, adjectives or adjuncts; question words, which may be nouns or adjuncts; and ideophones, which are phonologically unusual and generally act as coverbs or adjuncts. Notably, some word classes are very small or absent compared to the equivalents in most languages: in particular, verbs and adverbs are closed and relatively small, and there are no adpositions. A small class of conjunctions might be named, but would consist only of the single word naa ‘and’.

[Confusingly, I’ve been using ‘adjunct’ as the name of a word class, but ‘verbal adjunct’ as the name of an unrelated syntactic slot and associated construction. Feel free to suggest better terms if you can think of any!]

Closed word classes

By far the largest of the closed word classes are the verbs. Proto-‘Savanna’ has a limited set of ~150 verbal roots: there are no derivational processes to form roots from other word classes, nor are loanwords accepted. (Loaned verbs tend to become coverbs instead.) Verbs are defined as a word class by taking the verbal extension suffixes, requiring a subject, and being able to occur after reduced subject pronouns and the verb complex particles. Perhaps the most noteworthy verbal feature, however, is the alternation between multiple roots. Most verbs have two distinct roots: one used in imperfective contexts, the other used in perfective contexts. There is often an obvious synchronic relationship between the two roots, but the form of one root cannot be predicted from the other. For this reason, when citing verbs I give both the imperfective and perfective roots (in that order).

Verbs can be further subdivided into intransitive, transitive and ambitransitive classes. Intransitive verbs like qefaa~qefay ‘rise’, phage~phaŋi ‘break’ can take only a subject, whereas transitive verbs like siwe~siwi ‘have’, walha~waalhi ‘go’ require both a subject and an object. Ambitransitive verbs, which can occur both with and without an object, are rarer; most are agentive, as with waq~mah ‘do, happen’, though some are patientive, like segek~segeg ‘die, kill’. There are no true ditransitives.

Another verbal subclass of note is that of the generic verbs. This refers to a set of verbs with highly general and polysemous semantics. Most languages have at least a handful of generic verbs, but in Proto-‘Savanna’ they are of particular importance: all of the most frequent verbs encountered in general conversation are generic, and only some generic verbs may take verbal adjuncts. The following are a few examples of frequent generic verbs:

waq~mah ‘do’ (ambi. s=a): do, happen, make, create, cause, exist (adjunct-taking)
fawetl~fawetli ‘say’ (intr.): say, speak, talk, ask, shout, scream, laugh, cry, murmur, call (adjunct-taking)
siwe~siwi ‘have’ (tr.): have, get, take, hold, catch, carry, use, control, constrain, rule (adjunct-taking)
lhiise~lhiisi ‘perceive’ (tr.): sense, see, hear, smell, touch, taste, feel, understand, infer (adjunct-taking)
gadiq~gadye ‘absorb’ (tr.): absorb, contain, ingest, have something attached to, have something inserted in, eat, drink, conquer, have something carved on
ndeqis~ndisi ‘place’ (tr.): place, set down, hold tight, stabilise, settle, halt, comfort
fas~fasi ‘stand’ (tr.): be at/on/in a place, stand up, walk, reside, arrive at, leave

In practice, there are several mechanisms for disambiguation. A verb adjunct, if present, will usually have a highly specific meaning, thereby clarifying the relevant verb. Similar remarks can be made for serial verb constructions, as well as for the highly elaborated aspectual system. As with all languages, context also plays a large role in determining the intended meaning.

The other closed word classes are much smaller. There are six full personal pronouns:

Person	Singular	Plural
1	bal	bel
2	ndil	ndel
3	ŋay	ŋey

When used before a verb or in a possessive phrase, these have corresponding reduced forms:

Person	Singular	Plural
1	be	baa
2	ni	nde
3	qi	ŋii

(Note that I am somewhat inconsistent in how I write these pronouns. Before nouns, I tend to write them without a following space, i.e. as a prefix to the following noun; whereas in the verb complex, I follow them with a space. Given that this language is known from written sources and reconstruction, we have a dearth of phonological information, but grammatical behaviour suggests these act as genuine prefixes in the first case at least. If spaces turn out to be unhelpful, I might switch to using a hyphen for one or both of these usages.)

The demonstratives — often considered ‘pronouns’ too, though here the differences are obvious — are for the most part formed transparently, and may be summarised as follows:

	Proximate	Medial (visible)	Distal (non-visible)	Indefinite/interrogative
Attributive	si- / sii-	tse- / tsu-	ŋa- / ya-	naaŋ / maaŋ
Person	siqa / siiqa	tseqa / tsuqa	ŋaqa / yaqa	naaqa / maqa
Thing	sike / siike	tseke / tsuke	ŋake / yake	naak / maqek
Amount	sidaw	tsedaw	ŋadaw	naadu
Place	sil	tsel	ŋal	naal
Time	sitl	tsetl	ŋatl	naatl
Manner	siya	—	ŋaya	naaya

Some demonstratives have different forms for singular and plural nouns, and are shown with the singular form on the left and the plural form on the right (separated by a slash). The demonstratives cross-cut word classes: ‘person’ and ‘thing’ words are syntactically nouns, whereas ‘amount’ words act as adjectives, ‘time’ and ‘place’ words as adjuncts, and ‘manner’ words as adverbs. The attributive demonstratives are not independent words at all, but nominal prefixes: qetlaag thaŋ ‘the rock’ vs siqetlaag thaŋ ‘that rock’. However, naaŋ and maaŋ behave like adjectives, rather than being prefixes. Indefinite and interrogative words are not distinguished: naak phaŋi ‘something broke’ vs naak bey phaŋi ‘what broke?’.

Open word classes

By far the largest word class is that of nouns. These are distinguished from other nominals by the ability to take articles, be possessed, and be modified by other nouns and adjectives. Morphologically, nouns can take demonstrative and personal prefixes; derivational affixes forming adjectives or other nouns is also observed.

To a large extent, adjectives behave the same way as nouns: a single adjective forms a valid NP, and can take articles. (In these usecases, note that they refer to an ellipsed object with some quality, rather than the quality itself: daqat ‘big / big thing’, yenge ‘old / old person’.) However, adjectives are more restricted in their combinatorial possibilities than are nouns. Adjectives cannot be possessed, cannot be modified by nouns or relative clauses (though they may be modified by other adjectives), and cannot take nominal prefixes. However, nouns may be modified by multiple adjectives, whereas at most one nominal modifier is possible. Similarly, all adjectives may act as verbal adjuncts, whereas many nouns cannot appear in this position. Adjectives also have different derivational possibilities to nouns.

Closely related to the adjectives are the numerals 1–10 (which I supply for Janko):

1. yama
2. kheyi
3. quse
4. maŋaa
5. thawu
6. tuya
7. wagi
8. sulu
9. talhaa
10. ndatla

These are even more restricted than adjectives: they may not be modified at all, and may make up a standalone NP only when accompanied by the corresponding definite article.

The adverbs are a small but open class. There are only around ten monomorphemic adverbs, but the class is not entirely closed: any adjective may be converted to an adverb by reduplication. Adverbs are distinguished from adjuncts by their position immediately adjacent to the verb, and by adjectives by modifying the verb rather than the noun.

The final noun-like word class is that of the coverbs. Considered cross-linguistically, this is a highly unusual word class with distinctive distributional properties. Coverbs may head neither an NP nor a VP. Instead, they act as verbal adjuncts by occurring after a generic verb to alter its meaning. This is not a simple modifier–modified relationship: the verb+coverb combination acts as a unit, with its own argument structure and semantics. For instance:

lhiise meŋul lit. ‘perceive eye’ = ‘see’ (tr.)
fawetl yaqa lit. ‘say laugh’ = ‘laugh’ (intr.)

Despite their differing syntactic distribution, the line between coverbs and nouns is thin, if not entirely nonexistent. In particular, coverbs can look extremely similar to object nouns. For example: is the second word in gadiq tlaame ‘take.in food’ = ‘eat’ a coverb or a noun? It certainly seems like a coverb, and has similar semantics to other coverbs. However, it could just as easily be a simple VP, with tlaame being the object of the verb gadiq. A key test is whether the ‘coverb’ may be modified, since verb adjuncts resist modification. In this case, the test shows that tlaame is a noun: constructions like gadiq tlaame thaŋ ‘eat the food’ are perfectly acceptable. Another test is whether the verb+‘coverb’ combination can take an additional nominal object — if so, the second word must indeed be a coverb, since Proto-‘Savanna’ verbs are never ditransitive. This is how we know that meŋul is acting as a coverb in lhiise meŋul. Similarly, we know that yaqa in fawetl yaqa must be acting as a coverb, since fawetl is normally intransitive and cannot take a nominal object. Using such tests, words can often be assigned a coverbal or nominal identity. However, numerous words may be used in both word classes, with some being most commonly used as coverbs, others most commonly used as nouns, and many commonly used as both (as is the case for e.g. meŋul). Probably the best option is to treat nouns and coverbs as two ends of a graded continuum.

The last major open word class constitutes the ideophones. This class is syntactically heterogenous: most members can be considered coverbs or adjuncts (or both), but ideophonic adjectives and even nouns can also be found. Ideophones are instead defined by their distinctive phonological and semantic properties. Phonologically, ideophones often use phonemes rare in other words, such as /h tɬʰ ɡ uː/. Unlike other roots, many—perhaps most—ideophones are reduplicated. They also tend to have unique prosody and intonation. Semantically and pragmatically, ideophones impart a particular vividness to a text by dramatically communicating sounds, manners or ideas. Syntactically, as coverbs they occur as the complement of waq ‘do’ or walha ‘go’:

Qi waalhi haŋuhaŋu lit. ’it went haŋuhaŋu’ = ‘s/he/it blowed and blowed’
Qaŋeth thaŋ waq phuum fas sasay lit. ‘the person did phuum at home’ = ‘the person was nice and snug at home’

As manner adjuncts, they are placed mostly at the ends of sentences:

Be fasiŋu ŋake qefay ŋal tsektsek lit. ‘I stood up at that rising there tsektsek’ = ‘I suddenly stood up and appeared there’

Though not entirely unconstrained, the class of ideophones is highly open-ended. Speakers often innovate new ideophones as they speak. They are expressives, hence their distinct form and function from the rest of the lexicon.

[vegfarandi]

Very cool and thanks for the praise

Just a note regarding your "generic verbs"; from my research, there seem to be two methods for encoding sense verbs beyond a plain verb for each sense. Several languages that use Duriac's method of manipulation verb + body part (e.g. English "lay eyes on") and there are several that use "see" or "hear" + lower-sense noun (for instance, in Slavic languages, you tend to "hear smells"). Sight and hearing are considered "higher senses" for humans and are the most likely to have unique lexemes – taste, smell and touch are less likely. And there is an implicational hierarchy, a language only has a verb for taste/smell/sense by touch if there is a verb for see and hear. However, I don't think there's many languages where the basic sensation verb means "perceive" – I might be wrong, but it's most likely to, at least originally, mean "see" or "hear" vs. "perceive" – only because perception in the abstract isn't particularly useful in daily life.

[Vardelm]

Still reading through when I have time while working, but liking it so far. I have 1 nitpick and 1 correction for you.

	Labial	Alveolar	Lateral	Palatal	Velar	Glottal
Nasal	m ᵐb	n ⁿd			ŋ ᵑɡ
Voiceless stop	p pʰ	t tʰ			k kʰ	ʔ
Voiced stop	b	d
Affricate		t͡s t͡sʰ	t͡ɬ t͡ɬʰ
Fricative	f	s	ɬ		ɣ	h
Approximant			l	j	w

• The nasals exist at bilabial, alveolar and velar PoAs. All voiced stops can be prenasalised; in fact, the voiced velar stop exists only in its prenasalised form /ᵑɡ/.
• There are four stop series: voiceless, aspirated, prenasalised (already mentioned) and voiced. Generally the aspiration is strong, and voiced stops are fully voiced, occasionally slightly implosive. The glottal stop exists only in one series, conventionally taken to be voiceless; only three series of velar stop exist, as there is no fully voiced velar stop */ɡ/.

It seems a bit odd to me to list the pre-nasalized stops in the nasal row. I have only seen them listed as stops in other phonologies. You also mention them in your 2nd bullet point about stops.

Closed word classes

... Most verbs have two distinct roots: one used in imperfective contexts, the other used in imperfective contexts. ...

This sentence is imperfect. I assume one of those would be "perfective contexts".

[bradrn]

Just a note regarding your "generic verbs"; from my research, there seem to be two methods for encoding sense verbs beyond a plain verb for each sense. Several languages that use Duriac's method of manipulation verb + body part (e.g. English "lay eyes on") and there are several that use "see" or "hear" + lower-sense noun (for instance, in Slavic languages, you tend to "hear smells"). Sight and hearing are considered "higher senses" for humans and are the most likely to have unique lexemes – taste, smell and touch are less likely. And there is an implicational hierarchy, a language only has a verb for taste/smell/sense by touch if there is a verb for see and hear. However, I don't think there's many languages where the basic sensation verb means "perceive" – I might be wrong, but it's most likely to, at least originally, mean "see" or "hear" vs. "perceive" – only because perception in the abstract isn't particularly useful in daily life.

Kalam, which was my main inspiration for this language, uses a ‘perceive’ verb, e.g.: (Pawley 2011:64)

Yad

I

[tumuk

thunder

ag-e-k]

sound-DS.PAST

nŋ-n-k

perceive-1SG-PAST

‘I heard thunder.’

That being said, the term ‘perceive’ is basically an artefact of glossing: it’s just a generic ‘see/hear/feel/know’ verb, and it isn’t easy to find an English gloss for that. ‘Perceive’ certainly does carry a connotation of abstractness which isn’t really intended.

• The nasals exist at bilabial, alveolar and velar PoAs. All voiced stops can be prenasalised; in fact, the voiced velar stop exists only in its prenasalised form /ᵑɡ/.
• There are four stop series: voiceless, aspirated, prenasalised (already mentioned) and voiced. Generally the aspiration is strong, and voiced stops are fully voiced, occasionally slightly implosive. The glottal stop exists only in one series, conventionally taken to be voiceless; only three series of velar stop exist, as there is no fully voiced velar stop */ɡ/.

It seems a bit odd to me to list the pre-nasalized stops in the nasal row. I have only seen them listed as stops in other phonologies. You also mention them in your 2nd bullet point about stops.

I know it’s odd. I just wanted some way to avoid using an excessive amount of rows, and they seemed to fit with the nasals best. (Every nasal has a corresponding prenasalised stop; this isn’t the case with the plain voiced stops.)

Closed word classes

... Most verbs have two distinct roots: one used in imperfective contexts, the other used in imperfective contexts. ...

This sentence is imperfect. I assume one of those would be "perfective contexts".

Whoops! Fixed.

[Vardelm]

Got through the rest. My only other comment is that I like the generic verb bit (so +rep to vegfarandi as well ). I haven't done a ton of work on the "semantic organization" for my langs yet, so it's compelling to see it in others.

[bradrn]

Got through the rest. My only other comment is that I like the generic verb bit (so +rep to vegfarandi as well ). I haven't done a ton of work on the "semantic organization" for my langs yet, so it's compelling to see it in others.

Thank you! That was easily the most difficult part of the post to figure out — the previous version of this language didn’t have any coherent semantics around its generic verbs, so I had to basically re-create them from scratch. And I’m terrible at lexicon-building from scratch. It did give me the opportunity to learn about the Natural Semantic Metalanguage, though, which was powerful enough as a tool that it made up for the effort. (Wierzbicka’s Cross-Cultural Pragmatics: The Semantics of Human Interaction, though only tangentially related, and though I only skimmed it, can reasonably be said to have changed how I thought about semantics, pragmatics and culture.)

[vegfarandi]

Just a note regarding your "generic verbs"; from my research, there seem to be two methods for encoding sense verbs beyond a plain verb for each sense. Several languages that use Duriac's method of manipulation verb + body part (e.g. English "lay eyes on") and there are several that use "see" or "hear" + lower-sense noun (for instance, in Slavic languages, you tend to "hear smells"). Sight and hearing are considered "higher senses" for humans and are the most likely to have unique lexemes – taste, smell and touch are less likely. And there is an implicational hierarchy, a language only has a verb for taste/smell/sense by touch if there is a verb for see and hear. However, I don't think there's many languages where the basic sensation verb means "perceive" – I might be wrong, but it's most likely to, at least originally, mean "see" or "hear" vs. "perceive" – only because perception in the abstract isn't particularly useful in daily life.

Kalam, which was my main inspiration for this language, uses a ‘perceive’ verb, e.g.: (Pawley 2011:64)

Yad

I

[tumuk

thunder

ag-e-k]

sound-DS.PAST

nŋ-n-k

perceive-1SG-PAST

‘I heard thunder.’

That being said, the term ‘perceive’ is basically an artefact of glossing: it’s just a generic ‘see/hear/feel/know’ verb, and it isn’t easy to find an English gloss for that. ‘Perceive’ certainly does carry a connotation of abstractness which isn’t really intended.

Cool, I'll take a look at Kalam! And totally makes sense! Glossing is indeed an inaccurate art.

[Ares Land]

Very interesting! I'm eager to see more.

How are quantifiers handled (many, none, a few, etc?) Are datives and 'additional' objects marked?

[bradrn]

How are quantifiers handled (many, none, a few, etc?)

I haven’t quite worked that out… the previous version of this language had them working like numerals, but then again that means they would co-occur with articles, and I’m not quite sure if that makes any sense. I need to do some research into how natlangs handle these cases.

Are datives and 'additional' objects marked?

All verbs are at most transitive, so they don’t really need to be. Instead, they’re introduced via SVCs:

Be

1s

siwi

take.PFV

kumay

pot

thaŋ

DEF.SG

paatli

give.PFV

ndil.

2s

I gave you the pot.

Be

1s

waq

do.IMPF

sasay

home

thaŋ

DEF.SG

gadiq

absorb

segaag

frond

tleŋ.

DEF.PL

I’m attaching palm leaves to the house [for roofing].

I will say much (much, much…) more about this in a later post.

[Ares Land]

All verbs are at most transitive, so they don’t really need to be. Instead, they’re introduced via SVCs:

Ah, I should've guessed it

[bradrn]

Basic clause structure

Basic verbal clauses have SVO word order, or just SV for intransitive verbs:

[Kumay

pot

thaŋ]_S

DEF.SG

phaŋi_V.

break.PFV

The pot broke.

Qi_S

3s

fas_V

stand.IPFV

[ŋay

3s

ngen

POSS

sasay

home

thaŋ]_O.

DEF.SG

He is at his house.

Though this can be non-obvious when complex predicates are involved:

Be_S

1s

[fawetli

speak.PFV

yaqa]_V.

laugh

I laughed.

Qi_S

3s

waq_V

do.IMPF

[dise

yam

thaŋ]_O

DEF.SG

gadiq_V

absorb

nawun_O.

water

He is soaking the yam in water.

In the examples above, note that a pronominal subject is usually given in its reduced form. This is not required, though the full pronoun has something of an emphatic connotation: Be fawetli yaqa ‘I laughed’ vs Bal fawetli yaqa ‘I myself laughed’. A reduced pronoun may be used when a pronoun constitutes the entire NP, and is immediately before the VP. The subject may be left off entirely to form an imperative, which often additionally contains the final particle kaq:

Walha

go.IMPF

siwe

take.IMPF

nawun

water

tlii

INDF.PL

yusaa

come.IMPF

paatli

give.IMPF

bel

1p

(kaq).

IMP

Go and get us some water.

There also exist non-verbal clauses. These denote statements of identity, equality, presentation and adjectival predication. These may be formed by simple juxtaposition:

Ŋay

3s

yenge.

old

He’s old.

Danaq

3s

ganit.

woman

Danaq is a woman.

Especially in clauses of equation and presentation, a copular strategy can also be used, in which the particle nii is inserted after the first noun:

Tseke

that.VIS

nii

FOC

phingis.

palm

That’s a palm tree.

Despite appearances, nii is not a verb: most notably, it cannot take verbal morphology. It is probably best analysed as a focus particle rather than any sort of verbal element. Like the non-copular construction, it also forbids pronominal reduction in the first NP.

Negation differs between verbal and non-verbal clauses. In the former, it is accomplished is by means of the particle maag, placed before the verb complex:

Welus

man

thaŋ

DEF.SG

maag

NEG

gadye

absorb.PFV

ya-tlaame

NVIS.PL-food

tleŋ,

DEF.PL

qi

3s

mbu

but

qaathan-li

fall.IMPF-CONT

danit.

sit

The man didn’t eat his food, but just sat there.

Qi

3s

maag

NEG

walha

go.IMPF

lhiise

perceive

meŋul

eye

ŋay.

3s

He’s not looking at it. (Not: *he’s not going and seeing it)

Non-verbal clauses use the negative particle ŋaŋ instead:

Ŋay

3s

ŋaŋ

FOC.NEG

yenge.

old

He’s not old.

Tseke

that.VIS

ŋaŋ

FOC.NEG

phingis.

palm

That’s not a palm tree.

Polar questions are formed using a dedicated particle bey. In most cases this particle is placed after the first word of the clause (though see below for exceptions):

Ni

2s

bey

Q

siwi

take.PFV

ŋay

3s

yusaye

come.PFV

paatli

give.PFV

ŋay?

3s

Have you given it to him/her?

Sasay

home

bey

Q

daa

also

ni

2s

siya

this.way

qisa-tsi

cut.IMPF-ITR

dise

yam

tlii?

INDF.PL

Do you also cut yams like this at home?

Content questions are formed similarly, using bey alongside one of the interrogative/indefinite demonstratives mentioned earlier. The demonstrative must be fronted, with its head noun if one exists:

Naaya

what.way

bey

Q

sasay

home

ni

2s

qisa-tsi

cut.IMPF-ITR

dise

yam

tlii?

INDF.PL

How do you cut yams at home?

Maaŋ

which

dise

yam

bey

Q

sasay

home

ni

2s

siya

this.way

qisa-tsi?

cut.IMPF-ITR

Which yams do you cut like this at home?

Fronting is not restricted to content questions. It is common as a more general focalising transformation too. Especially outside question sentences, the fronting/copula particle nii is often inserted, and is required when focalising subjects:

Si-maqu

this-place

thaŋ

DEF.SG

(nii)

(FOC)

yangat

bird

tlii

INDF.PL

thabu

densely

fas.

stand

In this area there are many birds.

Pathu

Pathu

nii

FOC

siwi

have.PFV

tsisa

desire

tsuke

those.VIS

tleŋ.

DEF.PL

It was Pathu who wanted those ones.

Negative focal sentences use the negative focaliser ŋaŋ instead:

Pathu

Pathu

ŋaŋ

FOC.NEG

siwi

have.PFV

tsisa

desire

tsuke

those.VIS

tleŋ.

DEF.PL

It was not Pathu who wanted those ones.

Cleft sentences, using headless relatives—often without their usual article—are also common:

Si-maqu

this-place

thaŋ

DEF.SG

nii

FOC

naak

what

yangat

bird

tlii

INDF.PL

thabu

densely

fas.

stand

This area is where there are many birds.

I return now to the subject of clausal particles. These occur in four groups, distinguished primarily by syntax. Most numerous are the adjuncts, which occur at the periphery of a clause. Adjuncts generally indicate location in time or space, or modality: sil ‘here’, ŋatl ‘at that (distant) time’, qanta ‘now’, sasay ‘(at) home’, gatlaq ‘certainly’. Ideophones may also act as adjuncts: siki ‘with a cracking noise’, lampala ‘with a nice smell’, phuum ‘snugly’.

Somewhat smaller is the class of adverbs. These occur immediately next to the verb complex (most commonly before, but also after), and generally specify manner: matsiŋ ‘carefully’, thabu ‘densely’, qaabew ‘forcefully’. Adjectives may also be reduplicated to form adverbs: dawa-dawa ‘well’, ŋuwe-ŋuwe ‘badly’.

Somewhat more syntactically interesting are the second-position (2P) particles. We have already seen one example of these: namely, the question particle bey. As the name suggests, 2P particles are usually (though not always) placed in the second position in the sentence, after the first word. The 2P particles are as follows:

Particle	Class	Gloss	Function
bey	B	question particle	see above
yana	B	only	limitation
mu	A	past	backgrounded past event
daa	A	indeed, also	current clause expands on or confirms previous clause
mbu	A	but, contrariwise	current clause contradicts or gives opposing view
san	A	because	current clause gives a motivation for previous clause
na	B	thus	current clause gives a result or purpose
ya	B	already, just	reduces perceived temporal distance between two events; indicates recently changed state
ba	B	still	enlarges perceived temporal distance between two events; indicates continuing state
di	B	can	ability
tewu	A	(they) say	reported speech; hearsay

Note that many 2P particles are most commonly used for clause linkage, to compensate for the paucity of conjunctions, though of course they are hardly restricted to this usecase:

Bal

1s

mu

PST

yusaye

come.PFV

witsi-ŋu

contact.PFV-TEL

ŋay

3s

fasi

stand.PFV

naak

what.SG

baa

3p

waq

do

thaŋ,

DEF.SG,

qi

3s

ya

already

waq

do

ŋay.

3s

Having agreed on what to do, he is now doing it.

Qalit

sun

tewu

they.say

thaŋ

DEF.SG

nii

FOC

baalaq

sky

thaŋ

DEF.SG

gadiq-ŋu,

absorb.IMPF-RES,

qi

3s

na

thus

maag

NEG

siwe

have.IMPF

tsisa

want

bel

3p

may.

MOD

They say the sun is a part of the sky, so it does not like us particularly much.

Ŋay

2s

daa

indeed

nii

FOC

yana

only

mah

do.PFV

kuwi.

disease

Yes, he was the only one who got sick.

As shown above, 2P particles are in fact not always the second word in the clause! The exact positioning rules differ between clitics, and on the basis of syntax they can be divided into two position classes. Class A particles are those which in almost all cases may simply appear after the first word. Class B particles are more restricted: they may not follow an adjunct or appear within a focussed NP, and are moved forward to prevent this. Neither type of particle may interrupt a close-knit phrase such as noun compounds, and both skip over left-dislocated topics. Multiple clitics may occur in the clause; when this occurs they follow the order given in the table above.

Finally, the final particles are restricted, as the name indicates, to the final position in the clause. Generally these have pragmatic or illocutionary force: ngiya ‘surprise’, kaq ‘imperative’, kube ‘new information’, lis ‘politeness marker’, may ‘moderator’, qaa ‘assertion’.