Zompist Bboard Again

vegfarandi wrote: ↑Wed Jan 12, 2022 4:20 pm Both coordination and subordination certainly exist in Duriac, but the boundaries aren't entirely obvious, especially to a learner coming from a Western background.

This is of course very common.

3) is similar to some Australian languages, including Dyirbal

Do you happen to have any references? (Other than the Dyirbal grammar.)

4) is similar to some South-American languages.

…and Papuan languages, especially TNG. Note that they can’t actually be ‘converbs’ if they’re cosubordinate, since converbs tend to be specifically subordinate (at least in Haspelmath’s usage, which I think is wise to follow); you could just call these medial verbs, as is the Papuan convention.

bradrn wrote: ↑Wed Jan 12, 2022 11:11 pm 3) is similar to some Australian languages, including Dyirbal

Do you happen to have any references? (Other than the Dyirbal grammar.)[/quote]
Unfortunately, not very good ones, I've mostly seen off-hand remarks in various places in discussion of Australian languages. And honestly, I might have to do a bit more work on this particular part, it may not work as well as it has for the fairly simple test cases I've put it through! We shall see.

4) is similar to some South-American languages.

…and Papuan languages, especially TNG. Note that they can’t actually be ‘converbs’ if they’re cosubordinate, since converbs tend to be specifically subordinate (at least in Haspelmath’s usage, which I think is wise to follow); you could just call these medial verbs, as is the Papuan convention.

Interesting! The plot continues to thicken on this point. I sort of wish these terms didn't include the "verb" part in them because that implies to me either a mostly static participle-like form or a type of verb, whereas these are categories that take full inflection.

[

Vardelm wrote: ↑Wed Jan 12, 2022 11:42 am

bradrn wrote: ↑Wed Jan 12, 2022 7:03 am

evmdbm wrote: ↑Wed Jan 12, 2022 5:39 am 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.

I'll tell you what shall I just direct you here
https://www.verduria.org/viewtopic.php?p=16244#p16244

Case flagging in Duriac

Many polysynthetic languages don't do a lot of case marking, especially for core referents. But there's probably a cognitive reason for this. Many of those languages happen to be V-first, meaning the information about referents is already given before they're explicitly named, reducing the need to flag case roles. Duriac is V-last so there's some utility in flagging case. However, Duriac has rules that seemingly eliminate explicit markers when they conform with what's called the ideal core argument situation.

I use the term flagging per Haspelmath (2019). Other conventions: S = subject of intransitive verb, A = subject of transitive, P = direct object, IO = indirect object, X = applied object, term = {S, A, P}, argument = {S, A, P, IO, X}, oblique argument = {IO, X}, adjunct = non-argument NPs in an adverbial case

In Duriac, an ideal situation involves a sentient referent acting upon an inanimate referent. So:

• sentient A is always unflagged
• likewise an inanimate P is always unflagged
• animate A is flagged if the P is [EDIT] sentient (i.e. higher rank)
• animate P is flagged if the A is [EDIT] animate or inanimate (i.e. equal or lower rank)
• S is always unflagged.

So we have two explicit case markers used for terms, -e for the ergative case when A is flagged and -ad for the accusative case when P is flagged. Then we have the unmarked form seen in all terms under certain circumstances. For now I've chosen to analyze A/P as syntactically conditioned allomorphs of the ergative and accusative cases, respectively, leaving an additional case, the direct case, for S (which just happens to be syncretic with those unflagged forms).

A dative case marker -za is available to sentient and animate referents serving as IO. Inanimate referents cannot serve as IO and do not have this form. In causatives of transitive verbs, where the non-causative A is typically demoted to IO, an inanimate referent must instead become an adjunct, sometimes in the ablative, sometimes involving mbrambe 'because of', see later. The dative is also used to mark X when the benefactive applicative voice is marked on the verb; when X is inanimate, use the accusative marker -ad instead (but in this case, we call it the allative case). X controlled by the other two applicatives are simply accusative, subject to the same deflagging if inanimate as P.

So those are our argument cases: direct, ergative, accusative and dative.

Then there's a single adnominal case, the genitive. This one doesn't have a marker of its own so it borrows from others. On sentient/inanimate possessees, that is the dative marker -za. On inanimates, again it's the -ad, but one that doesn't go away. If the possessor is inanimate, flagging the genitive becomes optional, so you can say simply hadāl ḫuḫahum 'the roof of the house' instead of the fully expressed hadālad ḫuḫahum.

Lastly, there are four adverbial cases. In general, sentient nouns are resistant to these and need the supportive morph -na before the case marker. This is the same as the anaphoric pronoun na.

The allative is marginal; only inanimate referents can take -ad for allative. Animate nouns can take the locative -ka/ak in the same conditions. Sentient referents on the other hand, have to be couched in inanimate nouns, such as dikejar 'direction' or a body-part noun. It means something like 'to', 'toward' or 'into'.

The locative has -ka or -ak (the latter if the preceding segment is a stop). It means something like 'at', 'in' or 'on'.

Both the locative and allative are unflagged for proper nouns of place, e.g. Kādar inep 'I went to Kadar', not *Kādarad inep.

The ablative has -ube. It means something like 'from', 'out of' or 'off'.

The instrumental has -amma and means something like 'with', 'using', 'by' or 'in an X way'.

The case markers (-e, -ad, -za, -ak/ka, -ube, -amma) are phrase-final clitics and attach to the final word of the NP. However, they behave as affixes in that they are subject to hiatus collapse and they affect stress. While the final word is usually the noun, it can also be a relative clause or a number of other dislocated modifiers. Sentient nouns with the supportive -na take -na directly, so it may be detached from the case marker if something is moved after the noun within the NP.

In addition to case markers, there are a handful of words best termed postpositions phonologically and prosodically behave as separate words. The most postposition-like words are da and dabu, 'with' and 'without', which take an object in the accusative (unflagged if inanimate). Unlike -za, for instance, they don't pull stress. Whereas mamanza 'to father' is stressed on the final syllable mamanzá wheras mamanád da 'with father'. So the combination of the presence of the accusative marker and the difference in stress behavior shows these are slightly different than the case markers.

Then there's the word mbrambe 'because of' which needs a dative with sentient/animate referents, but no marker if inanimate. It's clearly related to a noun ḫambran 'reason' in an abbreviated ablative case form. The abbreviated ablative ending -be and the missing referential index ḫa- are associated with satellites (more on those later). But unlike satellites, which stand in apposition with same-case phrases rather than controlling their case per se, mbrambe seems to be triggering genitive agreement – but to complicate matters, it drops the ending with inanimates (which does not happen with typical genitives), perhaps in analogy with the behavior seen with da/dabu. So the simplest thing is to treat mbrambe as another postposition.

So to summarize, there are nine syntactic cases (direct, ergative, accusative, dative, genitive, allative, locative, ablative, instrumental), six morphological case markers, and at least three postpositions. Core case markers are not expressed in ideal argument situations, locative/allative is not expressed with proper nouns of place, inanimate objects of postpositions are unflagged and genitives may go unexpressed with inanimates. Together these non-expressions of case are called the rules of parsimonious case marking because they are all semantically motivated and aligned with statistically common and likely uses.

Making a table so I can visually see the marking hierarchy. Hope it doesn't step on your toes!

			Patient
		Sentient	Animate	Inanimate
	Sentient	sentient A unflagged sentient P ???	sentient A unflagged animate P flagged	sentient A unflagged inanimate P unflagged
Agent	Animate	animate A unflagged sentient P ???	animate A unflagged animate P flagged	animate A flagged inanimate P unflagged
	inanimate	inanimate A ??? sentient P ???	inanimate A ??? animate P unflagged	inanimate A ??? inanimate P unflagged

So, questions that I thought of from doing that little exercise:

1) What happens to sentient Ps and inanimate As? Are those flagged because they are un-ideal?

2) I don't know yet if the animate flagging is what I would expect yet. I think if I know the answer to question 1, I could see it better. I know WHAT the rules are, just not whether it seems logical or not.

3) What's the semantic divide between sentients, animates, & inanimates? Is that coming in a future post?

Vardelm wrote: ↑Fri Jan 21, 2022 4:29 pm Making a table so I can visually see the marking hierarchy. Hope it doesn't step on your toes!

1) What happens to sentient Ps and inanimate As? Are those flagged because they are un-ideal?

Not at all. This made me realize typos which I've corrected. Animate A is flagged if the P is sentient, not inanimate; i.e. if the P is higher ranking. Animate P is flagged if the A is equal or lower ranking. To answer your first question. Yes, since they are non-ideal, they are flagged. Sentient P and inanimate A are always flagged.

2) I don't know yet if the animate flagging is what I would expect yet. I think if I know the answer to question 1, I could see it better. I know WHAT the rules are, just not whether it seems logical or not.

Here's an adjusted table based on yours:

			Patient
		Sentient	Animate	Inanimate
	Sentient	sentient A unflagged sentient P -ad	sentient A unflagged animate P unflagged	sentient A unflagged inanimate P unflagged
Agent	Animate	animate A -e sentient P -ad	animate A unflagged animate P -ad	animate A unflagged inanimate P unflagged
	inanimate	inanimate A -e sentient P -ad	inanimate A -e animate P -ad	inanimate A -e inanimate P unflagged

Does that help?

3) What's the semantic divide between sentients, animates, & inanimates? Is that coming in a future post?

I could definitely do a post on that. For now I'll say this:

Sentient: human individuals, divine/profane beings
Animate: some words for children, body parts, animals, some weather
Animate uncountable: human institutions, groups, collectives, swarms
Inanimate uncountable: deverbals (except instance nouns), states, liquids/powders etc., body fluids/tissue, plant mass/grains etc.
Inanimate: everything else

vegfarandi wrote: ↑Fri Jan 21, 2022 3:06 pm Case flagging in Duriac

Many polysynthetic languages… happen to be V-first, meaning the information about referents is already given before they're explicitly named, reducing the need to flag case roles.

Where did you see this? The only V-first polysynthetic languages are in Mesoamerica and the Pacific Northwest. Almost all other polysynthetic languages are straightforwardly V-last.

Instead, I’d say that most polysynthetic languages don’t need to mark case because everything is already marked on the head, so marking it again would be redundant.

So we have two explicit case markers used for terms, -e for the ergative case when A is flagged and -ad for the accusative case when P is flagged. Then we have the unmarked form seen in all terms under certain circumstances. For now I've chosen to analyze A/P as syntactically conditioned allomorphs of the ergative and accusative cases, respectively, leaving an additional case, the direct case, for S (which just happens to be syncretic with those unflagged forms).

I see no need to analyse the S/A/P form as a separate case. You could just say that case-marking is optional: a marker is present in such-and-such circumstances, and absent everywhere else.

vegfarandi wrote: ↑Fri Jan 21, 2022 5:14 pm Does that help?

Yep!

Interesting that you have tripartite, accusative, ergative, and direct alignment depending on relative animacy. Nifty!

vegfarandi wrote: ↑Fri Jan 21, 2022 5:14 pm For now I'll say this:

Sentient: human individuals, divine/profane beings
Animate: some words for children, body parts, animals, some weather
Animate uncountable: human institutions, groups, collectives, swarms
Inanimate uncountable: deverbals (except instance nouns), states, liquids/powders etc., body fluids/tissue, plant mass/grains etc.
Inanimate: everything else

That makes it pretty clear. I look forward to seeing what other detail you have.

bradrn wrote: ↑Fri Jan 21, 2022 5:40 pm I see no need to analyse the S/A/P form as a separate case. You could just say that case-marking is optional: a marker is present in such-and-such circumstances, and absent everywhere else.

Yeah, I guess it's not absolutely needed, but "direct" would fit the bill pretty well. I think I've seen it used in both direct & tripartite alignments, which seems like it works well here.

Noun classes – Base Semantics

As discussed in a prior post (which I recommend reading first), Duriac has 13 noun classes, and for most nouns, this class is explicitly reflected in the noun's index. Nouns can be roughly divided into three groups based on animacy, sentient, animate and inanimate, and the latter two further by countable and uncountable.

There are three nominalization prefixes used on verbs which need to be mentioned. There's s- which is semantically neutral, similar to an infinitive or a gerund, t- which means something like "the one who" and y- which is specifically used with adjectival verbs. These are called s-nom, t-nom and y-nom. S- and t-nom can occur unindexed where they take z- (I8) and d- (S2) agreement respectively. Y-noms must appear with a referential index and this causes a process called yodding which affects the indexes d, ḫ and z, which become j, j and ź respectively. S- and t-[/i] optionally occur with an index but this reduces their verbness and increases their nounness; certain verbal morphology becomes impossible; and the word is often perceived more concretely or more as a label of identity, e.g. tâḫuz 'one who steals' vs. mitâḫuz 'male thief'.

Vardelm wrote: ↑Fri Jan 21, 2022 7:58 pm vegfarandi wrote: ↑Fri Jan 21, 2022 7:14 pm
For now I'll say this:

Sentient: human individuals, divine/profane beings
Animate: some words for children, body parts, animals, some weather
Animate uncountable: human institutions, groups, collectives, swarms
Inanimate uncountable: deverbals (except instance nouns), states, liquids/powders etc., body fluids/tissue, plant mass/grains etc.
Inanimate: everything else
That makes it pretty clear. I look forward to seeing what other detail you have.

I'll now expand on this a bit. In general, Duriac nouns have concrete reference, it's averse to abstract nouns, although they do exist. Furthermore, the majority nouns are roots; complex stems are mostly found in deverbals, including occupational terms and tool terms.

Using Dixon’s semantic concrete categories (per his A Semantic Approach to English Grammar), this is how they map to the noun class system:

• Human/god/spirit/demon: These nouns belong to the two sentient classes, S1 and S2, including terms of occupation (deverbal), rank and kinship terms (which are unique in several ways). Social groups and institutions are (A3). S1 is mostly masculine, S2 is feminine and epicene.
• Other animate: These belong to A1 and A2. Greater animals tend to be A1, smaller A2. Groups of animals are A3. Small swarming animals such as ants and flies often have no singulative term, existing only in a collective A3 form. Some words have a countable and uncountable sense, having dual status as A2 and A3 nouns.
• Body parts: A1 (face and hands) and A2 (other). Sometimes the same root is used across the two classes, e.g. dideg 'finger' vs. ḫadeg 'toe'. Note these are the same in the plural, yadeg 'fingers and/or toes'.
• Flora: I1, e.g. dihaśur 'apple tree' but fruits are I4: zihaśur 'apple'.
• Celestial and Weather: mostly I3, some I5 (night sky), I6 (material/water units), I7 (mass).
• Environment: landscapes I2 (more or less), stone I4, air I3, water I6, uncountable I7.
• Artifacts: Usually I4, containers I5, food and cookware I3, flat things I2.

As I said, Duriac doesn't have a lot of abstract nouns but there are some:

• Time: I5 generally as well as for nighttime, I2 for daytime specifically.
• Place: Mostly I2 but others as well.
• Quantity: I6 or I8 (if uncountable).
• Variety: I1 includes general words of variety digulud ‘kind, type’, diśamhe 'line (segment)', although most terms for shapes are I2 ḫagazd ‘circle’, handful are I4 zitri ‘shape’.
• Language: These are pretty uniformly I5: mipaźid ‘language’, midmû ‘word’, mebbâz ‘sound’, although ḫazu ‘name’ is I3.
• Speech acts: These tend to stay verbs: -ul ‘call’, -bual 'say', -äni ‘ask’ – if nominalized and indexed, they receive an I5 prefix by analogy of the above line ite, but if using the unindexed s-nom, they are referred to by z I8.
• General abstract: (‘idea’, ‘unit’, ‘result’) these kinds of things do not tend to be nouns at all but if pressed they'd be nominalizations, either I6 if countable or I8 if uncountable.
• Activities: Likewise these are pretty much just rendered as verbs. They can be nominalized as countable instances: I6 or uncountable concepts/processes: I8. This is particularly seen as the P argument of specific verb types as we will see in a later post.

Many languages have nouns for states and properties but in Duriac, these are uniformly expressed as adjectival verbs, including such a thing as ‘war’ -kadēr. They can be nominalized on occasion, if the state is segmentable (and thus countable), it is I5 (i.e. time), if it continuous it is I8. All properties, when nominalized, are I8.

While Dixon's classes help us see what kinds of words exist as nouns in Duriac compared to English, the below cut of the semantic space is more native to the language:

• S1 – men, masculine beings
• S2 – women, feminine beings, (adult) people in general
• A1 – face, hands, large animal
• A2 – other body parts, small animal, weather types
• A3 – human collectives and groups, animal groups, swarms
• I1 – long things, plant things, smelly things, the front of things
• I2 – flat things, places, daytime
• I3 – air, heat, fire, food, cookware
• I4 – artifact (that can be held in the hand), uneven object, fruit
• I5 – speech, nighttime, darkness, deep things (e.g. containers); via derivation: large (often non-portable) artifacts
• I6 – water, wet, drink, body-related (not liquid or body tissue), eggs, body-relative directions, instance of action, countable deverbal, countable collective
• I7 – uncountable liquid, uncountable powder, body fluid or tissue, uncountable plant
• I8 – other uncountable, uncountable product, general deverbal, deadjectival or state

Noun classes – Intra-Class Derivations

So while the semantic framework above apply to roots, things get muddier with various derivational processes which can move words out of semantically expected classes.

The S1 and S2 categories are reciprocally labile; most nouns here can alternate freely between m(i)- and d(i)- for male and female, e.g. marû ‘god’ / darû ‘godess’.

Note, I've gone back and forth about the numbering of these two, nothing is meant by the masculine class S1 being 1 vs. 2; but it's convenient to put S2 right before A1 in the table to have the two d-indexed classes adjacent to one another.

So-called encasement nouns can be formed with the prefix u-. This is seen in many body part nouns where an internal organ is derived from it's external container, e.g. ḫakū ‘head’ → ḫuakū ‘brain’. In a couple of instances, there is also involves a shift from A2 to A1, ḫayuneb ‘chest’ → duyuneb ‘lung’.

When body part nouns are used non-literally, the go from being animate to inanimate but retain their singular index. So a word such as ḫakū 'head' can be used in a non-literal sense meaning 'top, government, control' in which case it goes from A2 with a plural in y- to I3 with a plural in ś-. Similarly, a class A1 noun can become a class I1 noun with a plural in b-, e.g. disēza A1 'hand' vs. I1 'tool'. In some cases, such words can then be made sentient by incorporating it within the copula -ut and using the t-nom: mitasēzât 'lackey, minion, assistant' (lit. one who is a tool<hand').

Diminutives and augmentatives are another field. For sentient nouns, the prefixes -ge- and -j(i)- are used for augmentatives and diminutives, respectively:
migedug, digeda – big brother, big sister
migâtig – giant
dijitilir – small imp

Animate and inanimate countable nouns use -ḫu- and -źe- for augmentatives and dimunitives, respectively. While a word of A1 becomes A2 when diminutive (remember: A2 tends to be smaller animals), A2 always remains A2, even when augmentative. If the base word has a pseudo-index (which is common for animates), the derivative will have a standard index.
ḫaźekayzû ’pony’ (A2) ← dikayzû ‘horse’ (A1)
ḫaźedimśib 'small cow' (A2) ← tidimśib 'cow'
ḫaḫudda ‘big fly’ (A2) ← âdda ‘fly’ (A2)

For inanimates, if the base noun is I4, the augmentative reflex becomes I5:
miḫube ‘boulder’ (I5) ← zibe ‘rock’ (I4)

And all inanimate diminutives are class I4:
ziźehadâl ‘cottage’ (I4) ← hadâl ‘house’ (I2)
ziźebe ‘pebble’ (I4) ← zibe ‘rock’ (I4)

Uncountable nouns (A3, I7, I8) cannot be made diminutive or augmentative.

Collectives are sometimes formed with -yan- which causes yodded indexes: dibil ‘goat’ → jambil ‘herd of goats’. All these are A3 if the base is sentient or animate, I7 if specifically about liquid, I8 otherwise.

Ethnographic nouns tend to follow a simple schema:
Demonym: Bidūr 'Duriacs, Giduri' (Sp) (also Midūr ‘male Duriac’, Didūr ‘female Duriac’)
Territory name: Gidūr – land of Gidūr (I2)
Language name: Midūr – Duriac language (I5)

When nouns are compounded, the specifying root precedes the specified, and the class of the specified's root is retained in the derivative:
migam ‘wall’ + ḫūz ‘market’ → ḫagamūz ‘village’ (not *migamūz)

That last one went into a lot of detail but my goal is to keep these a bit more bite-size and high level. I've promised posts on satellites (aka "relational nouns" aka "directionals" aka "place/path adverbials"), locuphoric indexes (aka "personal affixes") and subordinated predicates. I'm also interested in going into what I have on pragmatics and courtesies. Let me know if any of these are of particular interest or if there's something in particular that you'd like me to discuss!

bradrn wrote: ↑Fri Jan 21, 2022 5:40 pm

vegfarandi wrote: ↑Fri Jan 21, 2022 3:06 pm Case flagging in Duriac

Many polysynthetic languages… happen to be V-first, meaning the information about referents is already given before they're explicitly named, reducing the need to flag case roles.

Where did you see this? The only V-first polysynthetic languages are in Mesoamerica and the Pacific Northwest. Almost all other polysynthetic languages are straightforwardly V-last.

Instead, I’d say that most polysynthetic languages don’t need to mark case because everything is already marked on the head, so marking it again would be redundant.

My hypothesis (although I have not done an academic survey) is that V-first polysynthetics have no case marking and the ones that do are all V-last. This may or may not be related to the word order, but it's possible that there's a a cognitive/functional reason. And so in thinking about case marking for Duriac, that's how I approached it. Duriac is V-last, it does have some case marking of terms, but it's limited to non-ideal circumstances, where the semantic animacy alone is not enough to indicate the role assignment.

bradrn wrote:

vegfarandi wrote: So we have two explicit case markers used for terms, -e for the ergative case when A is flagged and -ad for the accusative case when P is flagged. Then we have the unmarked form seen in all terms under certain circumstances. For now I've chosen to analyze A/P as syntactically conditioned allomorphs of the ergative and accusative cases, respectively, leaving an additional case, the direct case, for S (which just happens to be syncretic with those unflagged forms).

I see no need to analyse the S/A/P form as a separate case. You could just say that case-marking is optional: a marker is present in such-and-such circumstances, and absent everywhere else.

Well, it's that I consider the unmarked S case different from the unmarked A and P cases. Based on argument marking on the verb, behavior with the middle voice and word order tendencies, it's likely that the unmarked A and P cases are "deflagged", i.e. syntactically they have ergative or accusative case but morphologically this case is unrealized in certain (the statistically most common) circumstances. Whereas the unmarked S case never shows any case marking under any circumstances, so there's no argument to be made that underlyingly it's either ergative or accusative, there's no evidence to suggest that. So therefore it must be a separate case, the direct case.

Phrasal Verbs

Duriac verbs are in one sense a closed class – new roots do not pop up and there are no denominal (or de-otherwise) derivation processes. However, Duriac has a lot of phrasal verbs. In Duriac, what we mean by phrasal verbs are those verbs that have a set complement, usually as P (direct object) or X (applied object) which together form a lexically distinct semantic unit. Most of these are used with incorporation most of the time, yielding a single phono-syntactic word. Another type of lexically dependent relationships are satellite verbs – verbs with lexicalized relationships with satellites. Those we'll discuss later.

The most commonly encountered phrasal verbs in Duriac are verbs of attention and perception. Nearly all of these are composed of a manipulation-verb root such as -nād 'put' or -gū 'take' and a relevant body part as direct object. For instance, 'see' is y-iz -nād, lit. 'put eyes'. 'I saw him' can be rendered as nîzmunād, lit. I put eyes on him. This breaks down as n-i-iz-m-u-nād; 1s-PERF-eye-S1-LOC.APPL-put.

Without incorporating the noun and without the applicative, the phrase iz menaka ninād is also possible; y-iz m-e-na-ka n-i-y-nād A1-eye S1-DEICT-ANAPH-LOC 1s-PERF-A1-put. You have the option of using NI but not the applicative: menaka nîznād or use the applicative but not NI: iz nidmunād. Four ways of essentially saying the same thing with slight nuance.

Other phrasal verbs of visual perception also involve the use of the noun (y)iz 'eyes': y-iz -mbyē, lit, 'give eye', means 'look' – i.e. purposefully move the eyes in order to see something and y-iz -dab, lit. 'hold eye', means 'watch' – i.e. keep the eyes fixed in a certain area to see something.

Equivalent phrases exist for the other senses, but the precise syntax varies. For example, ḫasēg 'ear' has -sēg-u-ḫe, lit. 'get in ear', i.e. hear -sēg-u-dāŋ, lit. 'recline on the ear', i.e. 'listen'. In both these cases the ear is the X and in the former case, that which is heard is the P.

But there is a pattern; the visual perception pattern aligns with the fact that vision is primarily active, the audio perception pattern aligns with the fact that hearing is not optional. Smell is a combination of the two: di-ŋū -nād 'put nose' is equivalent to the pattern for 'see' and means 'smell' in the sense 'I caught the smell of x' but -ŋū-ka-ŋaḫ 'lean on the nose' is similar to the pattern for 'listen' in the sense of 'sniff', 'try to pick up a scent'. And there's also di-ŋū -dab 'hold nose' much like the pattern for 'watch' but this is only used for dogs, wolves and animals with a stronger sense of smell.

Overview of Verbal Morphology
Arguments and Voices

Here's a quick overview of the categories marked on the Duriac verb. Since it's been a while since I did my post on aspects, there may be a little bit of repetition so you can follow along without having to reread (but I suggest going back and catching up if you're interested). Definitely check glance at my third content post about indexes – these are used for typical argument marking on the verb which is the focus of this post. For simplicity, in the following examples I'll limit use to indexes n- (1s), ś- (2s), m- (S1 'male'), ḫ- (noun class I2 'place') z- (classes I5 'small thing' or I8 'nominalization of a verb'), and only the imperfective -a- and perfective -i- aspects.

Transitivity Classes
There are four types of transitivity inherent to each verb:
- intransitive monovalent (vi.) S
- intransitive bivalent (vi.+) S + IO
- transitive bivalent (vt.) A + P
- transitive trivalent (vt.+) A + P + IO

Logically avalent verbs such as verbs of light and weather are vi. and include dummy indexes that semantically relate to the verb's meaning:
ḫ-a-ḫebe I3-IMPF-blow 'it's windy' or '(the wind) blows'

Terms and Arguments
I use term to collectively speak about the the arguments S (intransitive subject), A (agent-like subject) and P (patient-like direct object); other arguments include IO (indirect object) and X (applied object). Terms are marked in the term slots, T1 and T2. Intransitives typically T1 only (with a few exceptions where T2 is used instead).
nadrimb n-a-∅-drimb 1s-IMPF-∅-sleep 'I was asleep' (vi., T2 empty)
inep ∅-i-n-ep ∅-PERF-1s-go 'I went' (vi., VC root shape requires use of T2 instead of T1)

In transitives, whether the A or the P goes in T1 or T2 depends on their relative animacy hierarchy; the higher ranked term must go before T1.
nizgū n-i-z-gū 1s-PERF-I5-take 'I took it' (vt., T2 occupied)

If T2 is occupied by A, in the inverse marker -e- must follow.
nizegū n-i-z-e-gū 1s-PERF-I5-INV-take 'it took me' (vt., inverse voice)

The + varients include a mandatory IO. IO appears before T2 with a following -a- glossed as OBL:
niśazbyē n-i-ś-a-z-mbyē 1s-PERF-2s-OBL-I5-give 'I gave it to you'

Vi+ is a bit unfamiliar to many speakers of European languages but is commonly found in languages of the Caucasus. They tend to be verbs of experience and in them the IO is typically experiencer argument as in the first example – or else an indirect or involuntary participant as in the second example here:
zanatab z-a-n-a-tab I8-IMPF-1s-OBL-know 'I know it' (lit. 'it is known to me')
naśagalgal n-a-ś-a-galgal 1s-IMPF-1s-OBL-talk 'I talked at you'

The IO can be omitted in the latter type.

The final argument type is X which occurs when an applicative voice is on the verb. There are three applicative voices: profactive -a-, spatial -u-/-ak- and comitative -am-. X's slot follows the T2 slot:
nizḫunād n-i-z-ḫ-u-nād 1s-PERF-I5-I2-SPAT-put 'I put it on it'

Causatives
There are two causatives, direct -ē-/-eg- and indirect -ege- (adjectival verbs use the suffix -aḫ instead). The prefixes attach immediately to the verb stem followed by the applicative marker. They don't mark an argument type exclusive to them as the applicatives do, but instead use the existing arguments and shift them around. In intransitives, a T2 is simply added, for both vi. and vi.+. In transitives, the A → IO and if present, IO → X with a profactive applicative. The profactive -a- collapses with the causative forming -ē- (unchanged), -âg- or -âge-. Note that this use of the profactive precludes causative vt.+ verbs from being otherwise applicativized as all the argument slots are filled.

nigud 'I jumped' → śinēgud 'you forced me jump'
nizgū 'I took it' → śinazēgud 'you forced me to take it'
niśazbyē 'I gave it to you' → śinaśazēmbyē 'you forced me to give it to you'

Reductive Voices
The middle voice marker -īl- allows the omission of the A argument in transitives:
izīlgū 'it got taken'

It can be combined with the inverse marker for an antipassive, but still using T2:
inīlegū 'I took (something)'

It is also used to mark exceptional reflexivity between A and P:
ninīlulā 'I worship myself' (compare: nimulā 'I worship him', verb -ulā 'worship')

Note that in most cases, reflexive meanings are inherent to verbs and such verbs tend to be either regular transitives morphologically, or sometimes have the same argument in both term slots. Reflexivity involving IO or X do not require the middle marker and simply use a repetitive index.

Still to Come: More on Aspects and Converbs – plus everything on Moods, Modal Voices, Discourse Forms, Nominalization, Focalization, Incorporation and Verb-Embedding Verbs.

The Dative case, Oblique Verbal Argument, Possession and Kinship Terms

Here's five facts we've seen in prior posts that (surprisingly?) interrelate:

1. the sentience or animacy of nominals is grammaticalized into sentient, animate and inanimate
2. possessed nominals are head- and dependent marked; the possessor is case marked, the possessee receives a [coindex]+u- prefix before its referential index: nu- (my) mu- (his), du- (her) etc.
3. for certain kinship words, this is an infix <[index]a> after the first syllable, e.g. da<na>dan 'my mother'
4. the marking of possessors is animacy-dependent: sentient and animate nominals receive dative -za âtigza muhadāl 'the man's house'; inanimates receive accusative/allative -ad hadālad ḫuḫahum 'the house's roof' (or sometimes no marking: hadāl ḫuḫahum 'the house's roof')
5. the oblique argument on the verb, nestled right after the aspect marker and before T2, separated from it by an -a-, can be a lexical mandatory argument in vi.+ and vt.+ verbs or it can be the demoted subject (in some theoretical frameworks called a chômeur) in a causative transitive verb.

Expanding on e), an additional use for the OBL is marking the possessor of a term (= S/A/P). When this is available corresponds with dative case marking – the possessor must be sentient or animate. For instance, 'his mother has arrived' can be rendered straightforwardly as:
damadan sû idep da<ma>dan sû i-d-ep <S1>mother here PERF-S2-go

or the marking of the possessor can move to the verb like so:
dadan sû imadep dadan sû i-m-a-d-ep mother here PERF-S1-OBL-S2-go

I want to note that the origin of these irregular kinships possessives, where an infix is used instead of a prefix, is the fact that kinship terms used to be verbs. Dadan 'mother' and maman 'father' come from an old verb -an 'be a parent' – so their possessive form is diachronically speaking a possessive oblique.

The verb -ep 'go' is a great example because it demands T2 slot usage for the S in the basic voice because of its VC root shape and thus allows us to distinguish the possessive oblique from the profactive applicative, which is also marked with -[index]a-, but before the stem as opposed to before the T2 slot – it's presence further blocks the T2 switacharoo so the form looks quite different:
dadan sû dimâp 'the mother has arrived for him'

In non-VC intransitives, the possessive oblique and the profactive would be ambiguous.

Compare to the verb -būḫbuḫ 'bark':
nuduŋru dabūḫbuḫ 'my dog is barking'
duŋru danabūḫbuḫ 'my dog is barking'

The latter could also mean 'a/the dog is barking for me' if <na> is the profactive.

With transitives – there's the potential for ambiguity as to which term, A or P, the possessive oblique pertains to but it is generally obvious from context:

dinazaḫuz d-i-n-a-z-aḫuz S1-PERF-1s-OBL-I5-steal 'she stole my [thing]' or 'my [female relation] stole it'

Locuphors

We've covered Duriac's aliophoric indexes which directly reference its class system in some detail. Alongside these indexes exist locuphoric indexes which instead reference speech act participants. There are five locuphoric indexes:

	Singular	Plural
1st person	n-	k-
1st person inclusive		nd-
2nd person	ś-	h-

Locuphoric indexes appear:

• in possessive prefixes: kuhadāl 'our (not your) house'
• in verbal term and argument slots ihep 'you all went', ndizgū 'we all took it'

You cannot replace a referential index on a noun with a locuphoric, for instance *nibāl, *'I the king'. However, nouns can be coreferential with locuphorics within a clause. The mismatched indexes are bolded below:

mibāl men nahazāz 'I, the king, am offering you this'

This type of usage occurs much regularly than in English, especially in formal settings. An appropriate response would be:

śubitaŋgēl kizzem 'we, your servants, have accepted it'

In addition to indexes, there are locuphoric pronouns:
īn 'I', ūn 'you (sg.)', dila 'we (incl.), kama 'we (excl)', yâ 'you (pl.)

yâ, which derives from (y)ah is quite irregular, īn and ūn are irregular in the accusative, losing their n: iyad and uad. They're most commonly found in the dative or genitive with -za to reinforce an IO or possessor.

Activity vs. State

Verbs in any language can be split along multiple semantic dimensions. But we often take for granted the splits that are prevalent in our native languages and those related to them. I've tried my best to achieve a very different set of semantic groupings from the languages I know best. In one of the first posts, I delineated a few key areas of difference:

• Verbs of motion and rest tend to encode manner but not direction (so there's run vs. crawl, but not ascend vs. descend)
• Verbs don't tend to encode what Dixon calls the "manip" (so there's just hit, no punch or slash)
• The semantic groups handled by adjectives in English are handled by verbs in Duriac (= adjectival verbs or va.)
• Attention verbs tend to be phrasal, using the relevant body part noun (see this post)
• Lastly, there are barely any secondary verbs.

I'd like to add a sixth item:

• The difference between activity and state is rarely expressed lexically.

Per Comrie (1971), aktionsart is split into two broad categories: occurrences and states. In Duriac, most natural (non-achieved) states are covered by adjectival verbs (va.) and have a few morphological quirks, including a lack of typical causatives and instead using the suffix -aḫ. Their inceptive (-aḫ-/-ayda-) essentially means "become/became/turn(ed)…" and they have no perfective per se – but essentially, their inceptive is roughly perfective in meaning: napad 'I am clean' → naḫpad 'I became clean'.

Out of occurrences, semelfactives (verbs that describe instantaneous occurrences), cannot take the imperfective, inceptive or cessative -uś-/-ukti-. Using the frequentative stem extension -śi, semelfactives can be made made frequentative/atelic: -bil ‘swing (once) → -bilśi ‘swing back and forth’ – and thus able to take on imperfective, inceptive or cessative. What this means is you can't say *namepin 'he's hitting me' or *naydamepin 'he's starting to hit me' – these are nonsensical unless you slow down time enough to capture the moment of "hitting". Therefore the frequentative is required to describe something ongoing: namepinśi 'he's hitting me (repeatedly)'.

All other types of occurrence verbs, however, can refer to states when made imperfective. For instance -gū 'take' will mean 'hold': mizgū 'he took it' but mazgū 'he is/was holding it' (i.e. he has taken it and still has it) – compare to "he's taking it somewhere", the difference being that the stative sense requires some kind of adverbial in English whereas Duriac does not have the same requirement. The verb -dab which we've seen before, where I glossed it as 'hold', is less of a pair with -gū than 'take' is to 'hold' in English – it can only refer to parts of the body. Perhaps 'keep' would have been a better gloss.

In some cases, you will see semantically related states and occurrences as different lexical items. For instance, a few exhibit reduplication for the state: -ba 'die' vs. -baba 'be dead'. However, this is a minority of the lexicon. Note that -baba as an adjectival can then take the shortened inceptive to mean 'become dead': maḫbaba 'he became dead' = 'he died' whereas -ba as a semelfactive cannot take the imperfective and its frequentative equivalent -baśi 'keep dying repeatedly' is more or less nonsensical outside of video games.

vegfarandi wrote: ↑Wed May 18, 2022 3:56 pm Activity vs. State

Verbs in any language can be split along multiple semantic dimensions. But we often take for granted the splits that are prevalent in our native languages and those related to them. I've tried my best to achieve a very different set of semantic groupings from the languages I know best. In one of the first posts, I delineated a few key areas of difference:

• Verbs of motion and rest tend to encode manner but not direction (so there's run vs. crawl, but not ascend vs. descend)
• Verbs don't tend to encode what Dixon calls the "manip" (so there's just hit, no punch or slash)
• The semantic groups handled by adjectives in English are handled by verbs in Duriac (= adjectival verbs or va.)
• Attention verbs tend to be phrasal, using the relevant body part noun (see this post)
• Lastly, there are barely any secondary verbs.

I'd like to add a sixth item:

• The difference between activity and state is rarely expressed lexically.

Somehow I missed this the first time around, but that's really neat and well-thought out.

I still can't help but think of the fruit when I see this language.

Travis B. wrote: ↑Thu May 19, 2022 10:52 am I still can't help but think of the fruit when I see this language.

Is that a problem?

vegfarandi wrote: ↑Thu May 19, 2022 11:54 am

Travis B. wrote: ↑Thu May 19, 2022 10:52 am I still can't help but think of the fruit when I see this language.

Is that a problem?

Not at all!