Duriac Thread

[vegfarandi]

My goal is to publish a full grammar of Duriac this year. (Famous last words, I've never managed to publish a full grammar to my satisfaction. But I'm hopeful this time it will succeed.) In the meantime I'm going to publish some nuggets of information on the language that don't quite belong in the reference grammar but that flesh the language out in various ways.

First, a brief introduction. In much of para-Lovecraftian fiction, Duriac is the language believed to have been the original language of the Necronomicon. Through my research, I have found that is spoken still to this day by the peoples of the continent Gidūr, sometimes called Lemuria, off the coasts between the Arabian peninsula and eastern Africa. It is a land not readily accessible through typical modes of transport. Duriac is the language demons, spirits and ghosts most prefer to communicate in, so if you're wondering why your séances never work, it's probably because you don't know Duriac yet.

The arguably most interesting features of the language are:

• a noun class system, somewhat reminiscent of Bantu and Caucasian languages, that interacts with plurality
• for transitive clauses, a parsimonious differential case marking system where a term (S "intransitive subject", A "transitive subject" or P "direct object") is only marked if it deviates from the ideal of sentient = A, inanimate = P. Animate nouns fall between the two and their marking depends on the animacy of O.
• up to four arguments marked on the verb and noun incorporation
• no tense but rich aspect system (note; aspect is a morphological term, semanto-syntactically, Duriac aspects incorporate elements of modality, evidentiality and sequencing. Duriac moods are distinct only from aspect morphologically.)
• nearly no overt coordination between clauses and also fairly light overt subordination

I'm sure we'll touch on some of these and more in the posts to come. Lastly, here's a brief guide to phonology, just as a reference for what's to come:

Consonants:

	Labial	Dental	Postalveolar	Velar	Epiglottal
Stops	p	t	c	k
Voiced	b	d	j	g
Prenasalized	mb	nd	nj	ŋg
Nasals	m	n		ŋ
Fricatives		s	ś	ḫ	h
Voiced		z	ź
Liquids		l, r	y

Vowels:

a, e, i, u /ə ɛ ɪ ʊ/
ā, ē, ī, ū /ɐː æː iː uː/

When derived from a hiatus collapse, the long vowels are spelled with ˆ in the romanization: â, ê, î, û

Syllables:

Syllables have the structure (C)(C)V(C)(C).

In the onset, any consonant may occur on its own. There are two kinds of onset clusters possible:

• Stop + r: pr, br, mbr; tr, dr, ndr; kr, gr, ndr
• Labial stop or nasal, or trill + y: py, by, mby, my; ry

Note that {ty, ky} > c and {dy, gy} > j.

In the coda, any consonant except c, j, nj, ŋg and h may occur on its own. There are four kinds of coda clusters:

• Voiced fricative + voiced stop: zb, zd, zg; źb, źd, źg
• Liquid + voiced stop: lb, ld, lg; yb, yd, yg
• Liquid + nasal: lm, ln; ym, yn

[*] yr occurs in two words, gremuyr ‘treasure’ and zeyr ‘complaint’.

Stress:

On the heaviest syllable from the right.

• Light syllables are open and have a short vowel: {–V}
• Heavy syllable have a long vowel or are closed with a short vowel: {–V̄, –VC}
• Superheavy syllables have a short vowel followed by a consonant cluster or have a long vowel in a closed syllable: {–VCC, –V̄C, –V̄CC}.

If the final syllable is light but all other syllables are also light, that's where the stress falls. The heaviest syllable close to the right edge of the word will pull the stress leftward.

[Travis B.]

When I saw the name of this thread I couldn't help but think of durians...

[vegfarandi]

One thing I'm obsessing over right now is Duriac's verbs. In an effort to avoid simply relexing English or Icelandic, I've been diving into how the lexicons of different languages treat semantic groups of verbs. The most helpful source has been R.M. Dixon's A Semantic Approach to English Grammar which has a categorization system for verbs based on semantics.

His primary division is between primary and secondary verbs:

• Primary verbs can be the only verb in a clause. Primary-A verbs have concrete things as their subject and object if applicable such as run, sit, touch, give, eat, rain, conquer, appoint, use, obey,
• B-type primary verbs will have a complement clause as the subject, object, or both, e.g. see, think, decide, say, like, annoy.
• Secondary verbs appear with another verb within the same clause such as must, begin, try, hurry, dare, want, delay, force, help, seem, matter.

But applying this to Duriac has been super interesting. As it turns out, due to the major differences between English and Duriac, Dixon's system doesn't map neatly onto the way Duriac verbs work but it's still helpful as a tool to know what kinds of clausal constructions Duriac has to be able to deal with to be fully functioning.

A note on citing and glossing conventions. Noun classes are labeled S, A or I for sentient, animate or inanimate, and then numbered, so S1-S2, A1-A3, I1-18. Plurality is simply appended: A1.PL, although the sentient nouns are syncretic across S1 and S2 so S.PL is sufficient there. Nouns with a class prefix are cited with a hyphen between it and the stem, so m-azūl 'warlock'. Verbs are cited with a hyphen before the root, so -nād 'put'. When citing a verb with an incorporated noun root (IN), it is done thus: -sēz-rab 'lead' (lit. 'hand-hold'). When an applicative is involved, -a- is the general applicative, -ak-/-u- is the spatial applicative, -am- is the instrumental applicative. The causatives are: -ē- (direct) and -ege- (indirect).

Major differences between Duriac and English verbs include the following:

• Verbs of motion and rest tend to encode manner but not direction. In this sense they tend to be similar to the Germanic verb stock of run, crawl, swim, trot; but there are no verbs like the Latinate stock of descend, adjoin, enter etc. A special subtype of nouns called satellites are used to convey spatial information such as direction and location. More on those in a future post!
• Verbs don't tend to encode what Dixon calls the "manip" (tool used to affect something with) in a lexical way. So Duriac doesn't have a verb punch; instead one says -zūl-am-pin lit. 'fist-hit' (note the instrumental applicative -am-). Likewise there would be no verbs beach (i.e. 'strand s-g on a beach') or shelve (i.e. 'put s-g on a shelf'), these would need to be paraphrased.
• The semantic groups handled by adjectives in English are handled by verbs in Duriac. For semantic reasons, these will never be entirely the same as a super typical verb such as hit, touch or break and this seems to be the case in most languages where these semantic domains are handled by verbs. For instance, adjectival concepts aren't quite as temporal as verbs; they tend to be stative rather than discrete events or actions. This affects the logical possible tenses used with these verbs.
• Attention verbs tend to be phrasal, using the relevant body part noun. So 'see' is d-iz -nād lit. put eye (on)' or, with incorporation: -iz-nād. So at the most verbose, one can say: nuyiz meka ninād, nu-y-iz m-e-ka n-i-y-nād, my-A1.PL-eye S1-DET-LOC 1s-PERF-A1.PL-put or most compactly, simply: nîzmunād, n-i-iz-m-u-nād, 1s-PERF-eye-S1-SPAT-put. Also featuring diz 'eye' are d-iz -dab, lit. 'hold eye' = 'watch', and d-iz -mbyē, lit. 'give eye' = 'look'. Similar phrasal verbs exist for ears, fingers, tongue etc. Notably, the liver signifies the site of thought and verbs of thinking are the same as verbs of attention.
• Lastly, there are barely any secondary verbs – most things that are handled by a secondary verb in English are handled through verbal morphology in Duriac, such as modality and causation.

The last point deserves more unpacking. Because there are so few secondary verbs in Duriac, some expressions may seem less specific in Duriac than English. As noted in the glossing note above, Duriac has two causatives, direct -ē- and indirect -ege-. The direct causative implies physical manipulation or contact while the indirect implies some kind of mental or social influence and thus covers everything between encourage, let, permit, allow, and ensure.

So, a sentence such as "the functionary ensured the wall was built" can be translated as âbrig migam mimegenjar, lit. 'the functionary made it so that the wall was built' but this could also mean "he allowed it to be built". If you want to be very specific you can add something like muzirigutamma 'with his diligence' or ziŋiamma 'with his permission'. But there are cases when a secondary verb is necessary. If only the primary action is to be negated, for instance, in the sense of English prevent, stop, etc., a subordination strategy is used: âbrig migam timīlnjarśa mimaḫ, lit. 'he made the wall the one that wasn't built'. Timīlnjarśa here is an agentive nominalization (t-) in the middle voice (-īl-) and the negative polarity (-śa).

[vegfarandi]

When I saw the name of this thread I couldn't help but think of durians...

Durians are plentiful on Gidūr, and much like the fruit, it is a land variously described as one of immense sweetness, overpowering unpleasantness, or intense disgust.

[Vardelm]

a noun class system, somewhat reminiscent of Bantu and Caucasian languages, that interacts with plurality

Looking forward to this in particular.

Also looking forward to reading post #3 on the verb groupings when my brain is less mushy.

[So Haleza Grise]

This looks great, I definitely want to read more.

[Ares Land]

I'm looking forward to more. I'm particular fond of the phonotactics; the language sounds lovely.

[vegfarandi]

Noun Classes and Indexes

One thing I've struggled with nailing on is how best to describe the class system. The class system interacts with morphology, syntax, semantics and derivation. What I've settled on is a method that I hope captures most of that but certainly there are alternative approaches possible. I'll start with a brief summary of the elements, followed by a brief summary of the description in the full grammar document, and at the end outline some of the alternative routes I could have taken.

Morphologically, there are 8 distinct aliophoric indexes, 5 singular: m-, d-, ḫ-, z-, r-, 3 plural: b-, y- and ś-. The indexes appear in 9 possible singular–plural pairs: m-b, m-ś; d-b, d-y; ḫ-b, ḫ-ś, ḫ-y; z-ś; and r-y. In addition, nouns with ḫ-, z- and r- can be uncountable and uncountable nouns have no plural. So from a morphological perspective there are essentially 9 classes, with 3 additional subsets.

Duriac syntax is sensitive to animacy on several levels and nouns can be organized on an animacy hierarchy: sentient > animate > inanimate. Taking this into account. Only nouns in m-b or d-b can be sentient. Nouns in d-y and ḫ-y can be animate. Nouns in m-ś, ḫ-b, ḫ-ś, z-ś and r-y can be inanimate. But complicating matters, so can d-b nouns. For convenience then, we must split d-b nouns into two subsets, the sentient one and the inanimate one. We now have a total of 13 categories:

Animacy	Class	Singular	Plural
Sentient	S1	m-	b-
	S2	d-	b-
Animate	A1	d-	y-
	A2	ḫ-	y-
	A3	ḫ-	—
Inanimate	I1	d-	b-
	I2	ḫ-	b-
	I3	ḫ-	ś-
	I4	z-	ś-
	I5	m-	ś-
	I6	r-	y-
	I7	r-	—
	I8	z-	—

There is also a handful of plural-only nouns, which could suggest another class, but these nouns behave exactly the same as other plural nouns on the syntax level and there's typically semantic motivation for the lack of singular. Whereas uncountable nouns require different treatment with quantifiers and thus behave differently from other singular nouns, thus warranting the special treatment.

So the above is the system I use. If you see a gloss A2, you know it's an animate countable noun referred to with the index ḫ- in the singular, y-. The one downside is you may have to look up the plural.

On most nouns, the index occurs as a prefix reflecting its class. It will have an epenthetic -i- if the stem begins with a consonant, unless the index is ḫ- in which case the epenthetic is -a-:
mi-bāl 'king' (S1)
ḫa-biltum 'pot' (I3)
m-edu ‘room, chamber’ (I5, note: no epenthetic)

The index is also used for agreement cross-indexing with verbs and agreement with some modifiers:
madār mibāl magar 'the great king stands'

It can occur with possessive prefixes (du-mibāl 'her king'), relative verbs (magar mibāl 'king who stands') and main verbs (mibāl magar 'the king stands').

No system is perfect without imperfections. Some nouns have an incongruent index; one that does not reflect the noun's class. Typically this occurs only in one number, either the singular or the plural, the other being congruent:
zêga rebil ‘profane book’ (not *zebil), cf. pl. śêga śebil ‘profane books’

In particular, several S2 nouns have y- in place of b- in the plural:
dadan ‘mother’, yadan ‘mothers’

Determiners exhibit incongruent indexes in certain classes: ze mibāl 'this king'.

Lastly, a few nouns have pseudo-indexes that don't look like any of the canonical indexes, perhaps hold-overs from an earlier version of the system with more classes:
ki-mur ‘shadow(s)’ (I5, not *mi-mur)
n-ūl ‘root’ (I1, not *d-ūl)
The pseudo-indexes â- and ha- are in complimentary distribution, â- occurs if the following vowel is short:
âbrig ‘functionary’ but hadāl ‘house’

In addition to the aliophoric indexes described above, there are locuphoric indexes referencing the speaker and listener which can do slightly different things. More on those in a future post.

---

An alternative approach would have been to describe the system in terms of the index pairs associated, using only 9 true classes, and then label words as sen., an. or inan. + uncount. if applicable. This would have simplified finding the plural form, but in some ways the glossing burden. I think the above methodology is the best compromise between information density and ease of use.

One last note. Organizing the indexes by their singular index reveals that ḫ- is the most versatile, pairing with all plural indexes and occurring with uncountable nouns as well, while m- and d- each have one plural pairing besides b-. z- and r- are exclusively found with inanimates.

Sorting by the plural index reveals a certain pattern: each plural pattern is associated with 3 singular patterns each (although d-b is associated with two classes in the above analysis); likewise uncountable nouns come in 3 flavors.

What I find fascinating about conlanging is that it often feels like discovery rather than creation. A lot of the analysis above is made after I "created" the system and although the system was informed by what I've learned studying Bantu and Caucasian languages, I did not design any of the patterning consciously, it is something that was revealed post hoc.

[bradrn]

I really like this! More please!

[vegfarandi]

A Brief Intro to Duriac Verb Aspects & Clause Chaining

The Duriac verb is quite complex with about 15 affix slots around the root. To keep things simple, we'll deal with a minimal verb first: first term – aspect – second term – root. The second term applies only to transitive verbs. Why "first/second term" and not "subject" and "object"? Because the relative order is determined by animacy hierarchy and when the roles are non-ideal, an inverse marker needs to be applied. For now, we'll deal with only ideal situations where the first term references S or A, and the second term – if present, references P.

So, as we saw in my last post, mibāl magar means 'the king stands': m-a-gar, where m- is the S1 index, -a- is the imperfective and -gar is the verbal root 'stand'. A transitive example would be: miḫzaz 'he broke it': m- S1, -i- perfective, -ḫ- index referring to something of either class A3, I2 or I3, and -zaz the verbal root 'break'.

Now we've also seen the two most important aspects: the imperfective -a- and the perfective -i-. In addition to these, there are the habitual, delimitive, inceptive, cessative and optative. These seven are sometimes grouped together as finite aspects although this is not quite accurate. Then there's six more: prospective, terminative, quoaditive, noncumitive, sequential and simultaneous. These are the non-finite aspects which is a more accurate label.

Why is the label "finite" inaccurate? Because most of the finite aspects can in fact be used in a non-finite manner. As mentioned above, the delineation is quite blurry in Duriac, as so many things tend to be on the continent of Gidūr. But it's short and convenient term and we can explain its limits later.

Sidenote: Asyndeton is the Primary Coordination Strategy

Duriac mainly relies on asyndeton, i.e. the absence of a coordinator to string together coordinate constituents and clauses. This applies to nouns: dīzg mīzg sû biḫūy 'the daughter and the son reached here'. Here, we see the coordinate NPs dīzg 'daughter' and mīzg 'son' repeated one after the other in the same case and co-referenced as the term of the verb with the plural sentient index b-. (Duriac also has the additive suffix -(e)ś which can be used to emphasize the coordinate status of nominals: dīzgeś mīzgeś ‘the daughter as well as the son’.)

There is no adversative conjunction (‘but, yet’), such contrasts are understood from context and can be with an intonation change: Käbida mabeld, maḫi ‘Käbida is poor but happy’

However, there is a disjunctive: deg ‘or’. It is repeated at the end of every coordinand: Käbida deg mududes deg ‘Käbida or his wife?’

When the coordinands are verbs in the interrogative -kāy, deg is omitted: maḫikāy, marrazeyrzuśakāy ‘is he happy, or does he not complain?’

Framing Events and Clause Chaining

Framing events anchor a narrative as a significant point and these tend to have finite aspect, most commonly perfective, but any is possible. These are surrounded by background events. Very commonly, the sequential -u- is what we see. The framing event can come first or last or be in the middle. In this example, it comes last, which builds anticipation:

Käbida daham mudgaz, mudeyaś, ḫilimgal mudḫunād, dudesza midadgēl
'Käbida cut an onion, boiled it, put it in a bowl, and (finally) served it to the wife'

This typical order places focus on the result: the wife getting fed. If Käbida were only serving the wife food so he can go do something else more important, using the perfective at the beginning of the chain would make sense.

The iconic (chronological) order of events is typically followed. However, the prospective -iŋ(u)- aspect can be used to background previously completed actions.

Käbida daham midgaz, ḫūmb miŋuḫbar 'Käbida cut the onion, having washed his knife before'

Its sibling, the terminative -aneḫ-/-aŋḫe- signals something that would happen after, but it may be predicted at the time the framing event occurred:

Käbida daham midgaz, dudesza maŋḫedadgēl 'Käbida cut the onion, later serving it to the wife'

These kinds of non-finite aspects can only semantically relate to a finite aspect predicate. If we for instance dropped the note about the previously cleaned knife into the first chain sample, where the serving of the wife is in the perfective:

ḫūmb miŋuḫbar, Käbida daham mudgaz, mudeyaś, ḫilimgal mudḫunād, dudesza midadgēl
'having cleaned the knife, Käbida cut an onion, boiled it, put it in a bowl, and served it to the wife'

then one would have to relate the cleaned knife to the serving: you would understand that Käbida intended to use the knife on the wife. A logical ending would be: …tet duḫasum yuz '…as her final meal'.

If the knife-cleaning business is related to cutting the onion, cutting the onion must be finite, or else the cleaning of the onion is marked as sequential much as the other verbs.

Aspect Pairs

Lastly for thist installment, I want to briefly touch upon "aspect pairs" which produce specialized meanings. Notably there is no ability to distinguish implicative conditionals (i.e. if this, then invariably that) and when-clauses in Duriac. Both are produced by pairing imperfective with imperfective: gaguln, yalumbep ‘If/when it rains, they get wet’ (lit. ‘it rains, they get wet’). The habitual can be used the same way: narradezedêg, ardīlśu ‘if/when she keeps bothering me, she will keep getting hurt’.

Predictive conditionals (if this happens, it's that will (most likely) result) are formed with the optative and an unbounded finite aspect such as the imperfective or habitual. Counterfactual conditionals are the same way, but replacing the optative with the delimitive.

Others exist but I'll leave it at that for now.

[bradrn]

Ooh, a polysynthetic language with converbs and hierarchical alignment! I’m starting to really like this.

Now we've also seen the two most important aspects: the imperfective -a- and the perfective -i-. In addition to these, there are the habitual, delimitive, inceptive, cessative and optative. These seven are sometimes grouped together as finite aspects although this is not quite accurate. Then there's six more: prospective, terminative, quoaditive, noncumitive, sequential and simultaneous. These are the non-finite aspects which is a more accurate label.

Does ‘finite’ even have any sensible definition outside IE? I suggest calling the ‘non-finite aspects’ ‘converbs’ (since that’s what they are); the ‘finite aspects’ can then just be plain ‘aspects’.

[vegfarandi]

Ooh, a polysynthetic language with converbs and hierarchical alignment! I’m starting to really like this.

Now we've also seen the two most important aspects: the imperfective -a- and the perfective -i-. In addition to these, there are the habitual, delimitive, inceptive, cessative and optative. These seven are sometimes grouped together as finite aspects although this is not quite accurate. Then there's six more: prospective, terminative, quoaditive, noncumitive, sequential and simultaneous. These are the non-finite aspects which is a more accurate label.

Does ‘finite’ even have any sensible definition outside IE? I suggest calling the ‘non-finite aspects’ ‘converbs’ (since that’s what they are); the ‘finite aspects’ can then just be plain ‘aspects’.

That's a really good suggestion! Let me take a look at doing that; that actually gives me a whole new idea on how to approach the description of these.

[evmdbm]

Ooh, a polysynthetic language with converbs and hierarchical alignment! I’m starting to really like this.

What's a converb? He asked plaintively...

[bradrn]

Ooh, a polysynthetic language with converbs and hierarchical alignment! I’m starting to really like this.

What's a converb? He asked plaintively...

Quoth Haspelmath (1995, in Converbs in Cross-Linguistic Perspective):

[emphasis in original]
A converb is defined here as a nonfinite verb form whose main function is to mark adverbial subordination. Another way of putting it is that converbs are verbal adverbs, just like participles are verbal adjectives.

In English, converbs include clauses like Having got in the car, Jane drove to work, or Eve, being happy with her results, decided to continue research. In many other languages — which may or may not include Duriac, but will certainly include Hlʉ̂ as soon as I get around to writing the next post — converbs are used extensively for clause chaining: to adapt one of Haspelmath’s examples, in English this would look something like They, having reached the lake, having stopped the cart, having brought the suitcase, being at the bank of the lake, having opened the suitcase, took out the bottles. (In more normal English this would be rendered as: They reached the lake, stopped the cart, and brought the suitcase; being at the bank of the lake, they took out the bottles. Note that in English, converbs are still the most natural means of expressing simultaneity.)

[Moose-tache]

I always find it fascinating and confusing when languages have little overt subordination or clause-level coordination. This is really well implemented. Vel gert.

[bradrn]

I always find it fascinating and confusing when languages have little overt subordination or clause-level coordination.

Wait, what? Which languages have neither of those? As far as I understand, all methods of clause combination are either one or the other. Converbs are a prototypical example of subordination, and coordination is coordination even when it’s unmarked. There is a partial exception in the form of Papuan languages with medial verbs… though not a very good exception, given that their clause chaining is analysed as cosubordination, i.e. both coordination and subordination at the same time!

[evmdbm]

How can you be co-ordinate and subordinate at the same time? Although in fairness I think this might become something of a digression from Duriac grammar

I am looking forward to more Duriac and seeing the full grammar when it emerges. Incidentally I too agree that conlanging is discovery; there were some starling implications of my decision to "borrow" Japanese subject and object honorification for Vedreki. I hadn't expected that the auxiliaries I used for them would have to be reversed in the passive for instance

[bradrn]

How can you be co-ordinate and subordinate at the same time?

Basically, a cosubordinate clause is subordinate in that it’s dependent on a main clause, but coordinate in that it can’t be nested within that clause, i.e. it must be adjoined, not embedded. See Haspelmath for details.

I hadn't expected that the auxiliaries I used for them would have to be reversed in the passive for instance

How does that work?

[Vardelm]

Really enjoying what I'm seeing so far!

What I find fascinating about conlanging is that it often feels like discovery rather than creation. A lot of the analysis above is made after I "created" the system and although the system was informed by what I've learned studying Bantu and Caucasian languages, I did not design any of the patterning consciously, it is something that was revealed post hoc.

Yep, I love it when that happens! It can also just be an idea of how to do something different pops into your head, and even (especially?) if you first dismiss it, you soon realize that it makes so much more sense than what you had. Really cool feeling.

I hadn't expected that the auxiliaries I used for them would have to be reversed in the passive for instance

How does that work?

I'm interested in that as well, but might be better in a separate thread so this one isn't derailed from Duriac.

[vegfarandi]

Both coordination and subordination certainly exist in Duriac, but the boundaries aren't entirely obvious, especially to a learner coming from a Western background. Some of the features that make the boundary a little hazy-feeling include:

1. the near lack of coordinators (except for deg and if you count -(e)ś).
2. intransitive verbs (of which adjectival verbs are a subtype) can be relativized simply by placing them in front of a nominal. So as exemplified above, you can have madār mibāl 'great king' and mibāl madār 'king is great'. So when you have a clause madār mibāl magar, without proper intonation/punctuation, it's ambiguous between 'great king sits' and 'he's great, the king sits' because both would use the imperfective -a-. An alternative (and probably equally feasible) analysis is that the intransitive relative form is simply syncretic with the imperfective (i.e. identical to it).
3. the fact that some subordinate clauses have aspect (non-converb) forms such as conditionals: mibāl nēbḫētut, zigūgil nazgēr 'If I become king, I'll lower taxes'. This is in fact indistinguishable from 'Hopefully I become king. I'll lower taxes' and could be analyzed as such; i.e. that there are no conditionals per se.
4. all the clause chains involving the sequential -u- (as in Käbida's onion-cooking tale above) and simultaneous -uḫe- are examples of cosubordination as bradrn pointed out. The simultaneous and sequential converbs as I guess I'm referring to them now, are different in this sense from the other converbs (prospective, terminative, noncumitive, quoaditive) because those are all examples of straight up subordination.

From a typological perspective, 1) is similar to a lot of the Middle-Eastern ancient languages, Hebrew, Sumerian etc., 2) is similar to Japanese, 3) is similar to some Australian languages, including Dyirbal, 4) is similar to some South-American languages.