SOV help!

[Jonlang]

I stupidly decided that I would make one of my conlangs SOV for no reason other than "it's different to the others". This language is supposed to be kinda Latin-y in having "free word order" but its default is SOV, and I basically want it to be SOV but to have the freedom to change it to say SVO for poetry etc. It's also a literary form of the language - a more conservative version where the "modern" spoken form has moved to SVO. Because it is a literary language it's based on the grammar found in important religious/mythic texts with some additions and changes as time goes on, though not ones that reflect the spoken form (inspired by the situation with the literary and spoken forms of Welsh). So the literary language is kinda elitist and can have rules "just because" but nothing too extreme because it is also meant to be learnable without an expensive education (by the speakers of the colloquial form).

I'm having some trouble with word order, like when you get down into the finer points. So far what I have decided:

• Underlying SVO order with Adj-Noun order (usually), and Noun-Numeral order (I have no idea why, but it just feels good, I think it happened in a natlang so I felt justified.)
• Adjs only agree in number, not case, or person.
• Verbs agree in number, not person.

One thing I'm unsure of is how to stack verbs.... Take English "I want to go", do I make this "I to go want" as:

Code: Select all

pron.1st.SG.NOM go.INF want.PRES.SG ?

What about other clauses? Would "I know that he knows" stay the same:

Code: Select all

pron.1st.SG.NOM know.PRES.SG pron.REL pron.3rd.SG.NOM know.PRES.SG ?

What else should I be thinking about with an SOV language? I have no real experience of SOV languages or what feels natural to them. Also, anything I should be thinking about with regards to a language shifting from SOV to SVO would be appreciated.

[bradrn]

One thing I'm unsure of is how to stack verbs.... Take English "I want to go", do I make this "I to go want" as:

Code: Select all

pron.1st.SG.NOM go.INF want.PRES.SG ?

You could definitely do that! A lot of strictly verb-final languages put complements before the verb. But you could also postpose the subordinate clause, as in English — either one is possible, depending on what your language decides to do. My intuition here would be that if ‘to go’ is a single noun-like word, it would behave similarly to other nouns, but in general, since your language isn’t strictly V-final, there’s probably a bit more flexibility around subordinate clauses.

What about other clauses? Would "I know that he knows" stay the same:

Code: Select all

pron.1st.SG.NOM know.PRES.SG pron.REL pron.3rd.SG.NOM know.PRES.SG ?

Same story: it might stay the same, it might be arranged differently.

What else should I be thinking about with an SOV language? I have no real experience of SOV languages or what feels natural to them. Also, anything I should be thinking about with regards to a language shifting from SOV to SVO would be appreciated.

Honestly, I’m not sure there’s much to think about. Aside from the obvious word order correlates you’ve already mentioned, I don’t feel like word order itself has any particularly significant effects on the rest of the language. The only effect I can suggest with anything approaching confidence is that clause-chaining seems to happen more often in strictly verb-final languages, and even then I’m not convinced it’s a strong correlation.

(Stassen does have an intriguing article where he suggests that there’s a bunch of strong, global correlations between a whole set of typological parameters, including OV/VO word order status. I think he’s on to something, but it needs more investigation before I’d be confident in drawing any conclusions from it.)

[hwhatting]

Underlying SVO order with Adj-Noun order (usually), and Noun-Numeral order (I have no idea why, but it just feels good, I think it happened in a natlang so I felt justified.)

You mean underlying SOV, right?
And what is the SOV language that has noun-numeral?

One thing I'm unsure of is how to stack verbs.... Take English "I want to go", do I make this "I to go want" as:

Code: Select all

pron.1st.SG.NOM go.INF want.PRES.SG ?

Latin, which you mention, has both ire volo and volo ire, depending on various considerations (topic-comment, flow, looks good, fits the verse better...). More strictly SOV languages like Turkic systematically have infinitive plus finite verb, but actually often avoid infinitive + modal constructions by turning the modal into a suffix on the content verb, so go-VOL-(endings) instead of "want to go".

What about other clauses? Would "I know that he knows" stay the same:

Code: Select all

pron.1st.SG.NOM know.PRES.SG pron.REL pron.3rd.SG.NOM know.PRES.SG ?

Latin frequently has verb plus dependent sentence, but the more refined authors play around with this as well and like to build complicated puzzles out of the syntax. In the specific construction you mention, Latin uses an AcI: scio eum id scire or eum id scire scio " I know him to know it". Turkic languages would normally go for some gerund / nominalised construction like his it knowing know-I; they normally don't do dependent finite sentences except sometimes under Persian influence.

What else should I be thinking about with an SOV language? I have no real experience of SOV languages or what feels natural to them. Also, anything I should be thinking about with regards to a language shifting from SOV to SVO would be appreciated.

Short verson - if you go for a strict SOV language Turkic-type, you need to re-think almost everything and go for a lot of nominalised / gerund constructions. For something mildly underlying SOV like Latin, go with the flow, just keep in mind that an order that may look wrong or unnnatural to you, like go want-I or that he it knows know-I may not look that way for a speaker of such a language.

[bradrn]

More strictly SOV languages like Turkic systematically have infinitive plus finite verb, but actually often avoid infinitive + modal constructions by turning the modal into a suffix on the content verb, so go-VOL-(endings) instead of "want to go". … Short verson - if you go for a strict SOV language Turkic-type, you need to re-think almost everything and go for a lot of nominalised / gerund constructions.

I feel these recommendations are true only for one particular ‘type’ of SOV language with lots of clause-chaining, the kind which you term ‘Turkic-type’. (Other examples of this: the rest of ‘Altaic’, Trans–New Guinea, Ndu, NE Caucasian, maybe Quechua?) But there’s also many SOV languages which make much less use of non-finite forms: for instance, NW Caucasian, Skouic, Eskimo–Aleut, Ethiopian Semitic and Khoekhoe, to name just a few. And then there’s the SOV Sino–Tibetan languages which use nominalisations extensively, but (to my mind) quite differently to how Turkic uses them (e.g. Tibetan, Burmese, rGyalrongic). I can’t disagree that deranked forms are more common in SOV languages, but it’s not necessarily the case that Jonlang needs to ‘re-think almost everything’.

[linguistcat]

For a language that is underlying SOV but has free word order, agreement is your friend. Remember that free word order technically goes beyond just the positions of subject, object and verbs. It's good that you have some form of agreement between the nouns and adjectives since that can free up where adjectives can go while still connecting them to the correct nouns (or maybe some ambiguity could be part of the poetic tradition). Having Adj-N in an SOV language is pretty common AFAIK, but having numerals come after nouns in the same language is a little odd. Not saying it can't be done, especially with free word order, but it's less common especially with the default Adj-N construction.

I think any tendencies that are particularly strong in strict SOV langs could probably be played with but I would look into universals regarding how other aspects commonly line up. This is a link to a spreadsheet that I found with different word orders (including things like how nouns, adjectives and numerals can correlate). I'd like to say it was done by Artifexian, but I can't honestly remember to give proper credit. It is based more on whether you have a pre- or postpositional language, so keep that in mind.

[hwhatting]

but it’s not necessarily the case that Jonlang needs to ‘re-think almost everything’.

But that's where all the fun is

[bradrn]

For a language that is underlying SOV but has free word order, agreement is your friend.

Or case-marking, as in Australian languages. You just need some way of disambiguating which is subject and which is object.

but it’s not necessarily the case that Jonlang needs to ‘re-think almost everything’.

But that's where all the fun is

So I guess what I was saying is that he can re-think everything, but it doesn’t necessarily have to be in that one specific direction of ‘non-finite forms for everything’.

[zompist]

More strictly SOV languages like Turkic systematically have infinitive plus finite verb, but actually often avoid infinitive + modal constructions by turning the modal into a suffix on the content verb, so go-VOL-(endings) instead of "want to go". … Short verson - if you go for a strict SOV language Turkic-type, you need to re-think almost everything and go for a lot of nominalised / gerund constructions.

I feel these recommendations are true only for one particular ‘type’ of SOV language with lots of clause-chaining, the kind which you term ‘Turkic-type’. (Other examples of this: the rest of ‘Altaic’, Trans–New Guinea, Ndu, NE Caucasian, maybe Quechua?)

Quechua has a number of methods of combining verbs. The simplest is to use the nominalization or infinitive -y:

Mikuyta munani,
eat-nomn-acc want-1s
I want to eat.

Llamkayta qonqarun.
work-nomn-acc forget-past-3s
He forgot to work.

If the subjects of the verb differ, you use the nominalization -na instead:

Llamkananchiktam munanku.
work-nomn-1p-acc-evid want-3p
They want us to work.

There are a few more, mostly with adverbial meanings-- e.g. "he worked while whistling".

[bradrn]

More strictly SOV languages like Turkic systematically have infinitive plus finite verb, but actually often avoid infinitive + modal constructions by turning the modal into a suffix on the content verb, so go-VOL-(endings) instead of "want to go". … Short verson - if you go for a strict SOV language Turkic-type, you need to re-think almost everything and go for a lot of nominalised / gerund constructions.

I feel these recommendations are true only for one particular ‘type’ of SOV language with lots of clause-chaining, the kind which you term ‘Turkic-type’. (Other examples of this: the rest of ‘Altaic’, Trans–New Guinea, Ndu, NE Caucasian, maybe Quechua?)

Quechua has a number of methods of combining verbs. The simplest is to use the nominalization or infinitive -y … If the subjects of the verb differ, you use the nominalization -na instead …There are a few more, mostly with adverbial meanings-- e.g. "he worked while whistling".

OK, this does sound similar to Turkic indeed. But does Quechua use clause-chaining as extensively as Turkic?

[zompist]

OK, this does sound similar to Turkic indeed. But does Quechua use clause-chaining as extensively as Turkic?

I don't know Turkic.

My impression (but don't take this too seriously) is that there are a lot of things you can do in Quechua, but that doesn't mean they're common. I have a book of Quechua folk tales and I think the verbs are by no means the monsters you can find in a grammar.

[bradrn]

OK, this does sound similar to Turkic indeed. But does Quechua use clause-chaining as extensively as Turkic?

I don't know Turkic.

Emphasis was on the ‘clause-chaining’, not the ‘Turkic’… I don’t know Turkic either! By ‘clause-chaining’ I’m referring to the kinds of constructions covered in, say, https://www.sil.org/resources/publications/entry/7858.

My impression (but don't take this too seriously) is that there are a lot of things you can do in Quechua, but that doesn't mean they're common.

Of course, and this is the case in any language. But in this case I do happen to be interested in what’s common. (You can do clause-chaining in English too, but it’s thoroughly unidiomatic.)

[Imralu]

What about other clauses? Would "I know that he knows" stay the same:

Code: Select all

pron.1st.SG.NOM know.PRES.SG pron.REL pron.3rd.SG.NOM know.PRES.SG ?

This is off topic, but I wanted to point out that.COMP that’s not the role of a relative pronoun but a complementiser. Some languages do conflate them, such as English with that and (many of) the Romance languages with que/che, so of course it’s totally acceptable to do this, but a lot of people who.REL only speak English and a Romance language don’t realise that.COMP these are separate roles. None of the languages (Ø.REL) I’ve learned aside from English or romance languages uses the same word. (Although German only really orthographically distinguishes the complementiser dass from one type of relative pronoun in the neuter nominative and accusative, das.) I don’t know a single non-Western European language that treats them similarly.

[Torco]

I stupidly decided that I would make one of my conlangs SOV for no reason other than "it's different to the others"

my brother in arms you are

my first instinct, being currently working on a rather isolating SOVlang, is to string the verbs together with particles. this is probably no good, since this lang seems to be rather dense in inflections and so on. another way is something like object1 verb1 object2 verb2: you know, stuff like there go want for to want to go there (the person being somewhere in the verb, presumably), or even more complicated series, like here come this explain she want or something, as in, apparently, amharic

i found a cool article about how assamese does it, which is sov, and apparently they do it with conjunctions.

and I lost the link, but an argentinian dude had an interesting paper about the akkadian verb to go having being used in a stringing-verbs-together kind of way.

[bradrn]

I stupidly decided that I would make one of my conlangs SOV for no reason other than "it's different to the others"

my brother in arms you are

my first instinct, being currently working on a rather isolating SOVlang, is to string the verbs together with particles. this is probably no good, since this lang seems to be rather dense in inflections and so on. another way is something like object1 verb1 object2 verb2: you know, stuff like there go want for to want to go there (the person being somewhere in the verb, presumably), or even more complicated series, like here come this explain she want or something, as in, apparently, amharic

i found a cool article about how assamese does it, which is sov, and apparently they do it with conjunctions.

and I lost the link, but an argentinian dude had an interesting paper about the akkadian verb to go having being used in a stringing-verbs-together kind of way.

You’re conflating two quite different kinds of constructions here. The first is the ‘serial verb construction’, which is when one clause contains multiple verbs without formal linkers. (This is what Assamese has — I’m not sure where you saw ‘conjunctions’ there.) These generally are highly integrated: they’re essentially describing a single event by means of its components. For more on the subject, I can recommend ed. Senft’s Serial verb constructions in Austronesian and Papuan languages, and ed. Aikhenvald and Dixon’s Serial Verb Constructions (2006).

The second is ‘clause-chaining’, as in Oromo (not Amharic!). This is what I was talking about above, when a language strings lots of (usually subordinate) clauses together to form a narrative. The link I gave zompist earlier seems like a good overview here. In some cases verb serialisation and clause-chaining can have similar effects (e.g. in Kalam and Kilivila), but not often: verb serialisation is inherently a clause-internal operation, whereas clause-chaining involves stringing multiple clauses together. For some reason it seems like verb serialisation is most common in SVO languages and clause-chaining is most common in SOV languages, but this is merely a tendency.

[Jonlang]

Am I right in thinking that an SOV language generally has an order like Subject - Object - Object's verb - Subject's verb, i.e. "I see her come" as "I her come see"? Or "I need her to see" as "I (for) her to see need"?

[bradrn]

Am I right in thinking that an SOV language generally has an order like Subject - Object - Object's verb - Subject's verb, i.e. "I see her come" as "I her come see"? Or "I need her to see" as "I (for) her to see need"?

I’m not quite sure what you mean by ‘object’s verb’ and ‘subject’s verb’. These are just different kinds of complement clauses, and like all complement clauses, how they work is very highly language-dependent. (For instance, ‘I see her come’ is IIRC an English construction usually termed ‘secondary predication’: other languages might use verb serialisation or a finite verb to express the same construction.) That being said, neither of your examples would be implausible for an SOV language.

[Jonlang]

Am I right in thinking that an SOV language generally has an order like Subject - Object - Object's verb - Subject's verb, i.e. "I see her come" as "I her come see"? Or "I need her to see" as "I (for) her to see need"?

I’m not quite sure what you mean by ‘object’s verb’ and ‘subject’s verb’. These are just different kinds of complement clauses, and like all complement clauses, how they work is very highly language-dependent. (For instance, ‘I see her come’ is IIRC an English construction usually termed ‘secondary predication’: other languages might use verb serialisation or a finite verb to express the same construction.) That being said, neither of your examples would be implausible for an SOV language.

I'm very tired* and I don't know what the actual terms are or how to properly describe them, but "I see her come" is, as you can see, the subject seeing the object doing something, hence "subject verb" and "object verb" in my description. But "I see that she comes", making "her" the subject of a dependant clause could also work, I think. The problem I'm having here is making sure the final verb and initial subject agree, which they should in SOV, and not confusing them with any other verbs concerned with objects. "I threw the ball to John, but Kate caught it" would, I think, be:

Code: Select all

pron.1st.SG.NOM ball.SG.ACC J.DAT throw.SG.PST, but K.NOM pron.3rd.ACC catch.SG.PST

which seems simple enough because there's only one verb in each clause. But consider: "I saw John coming with a ball, but Kate saw John throw it". For these I see them as the following parts: I saw (sub) / John coming (obj) / a ball (obj) // Kate saw (subj) / John throw (obj) / it (the ball) (obj). So, my mind says this ought to be:

Code: Select all

pron.1st.SG.NOM J.ACC ball.SG.COM come.SG.PRES see.SG.PST, but K.NOM J.ACC pron.3rd.SG throw.SG.PRES see.SG.PST

but I'm not 100% sure. I'm not sure if 'it' should be marked as anything in the second clause.

* Too much to do and too little time I'm afraid.

[bradrn]

I'm very tired* and I don't know what the actual terms are or how to properly describe them, but "I see her come" is, as you can see, the subject seeing the object doing something, hence "subject verb" and "object verb" in my description. But "I see that she comes", making "her" the subject of a dependant clause could also work, I think.

I’d say that both constructions are complement clauses, just of different types — note that English has an unusually large and varied inventory of complement clauses compared to most languages (SOV or not). In ‘I see [that she comes]’, the complement clause is fully finite and introduced with ‘that’, whereas in ‘I see her [come]’ it is a so-called bare infinitive, analogous in surface structure to (say) ‘I want her [to come]’.

The problem I'm having here is making sure the final verb and initial subject agree, which they should in SOV

I don’t see what this has to do with SOV in any way. Why should agreement in SOV languages work any differently to agreement in any other language?

… and not confusing them with any other verbs concerned with objects.

I’m not sure what you mean by this.

"I threw the ball to John, but Kate caught it" would, I think, be:

Code: Select all

pron.1st.SG.NOM ball.SG.ACC J.DAT throw.SG.PST, but K.NOM pron.3rd.ACC catch.SG.PST

Yes, that seems reasonable.

But consider: "I saw John coming with a ball, but Kate saw John throw it". For these I see them as the following parts: I saw (sub) / John coming (obj) / a ball (obj) // Kate saw (subj) / John throw (obj) / it (the ball) (obj).

To me this feels like you’re ignoring the tree structure of the sentence. I’d write it as:

I saw John [coming [with a ball]_PP]_{secondary predicate}, but Kate saw John [throw it]_subordinate.

(Note: I saw John coming is a very different construction to I saw John come! They have to be, since ‘coming’ is functionally an adjective or noun, whereas ‘come’ is clearly a verb. I’m not sure how valid my analysis is of the former as being an instance of ‘secondary predication’, but it seems to line up with the usual usage of that term.)

How this gets translated into your language would depend on how it deals with the various adjuncts — which, by the way, has very little to do with whether it’s SOV or SVO. If it treats them as sentence-like dependent clauses, it might look like:

I-NOM [that John-NOM ball-COM is.coming.3s] saw.1s, but Kate-NOM [that John-NOM it-ACC threw.3s] saw.3s

On the other hand, if it treated them as nominalisations, it might look more like:

I-NOM [John’s ball-COM come-ing]-ACC saw.1s, but Kate-NOM [John’s throw-ing]-ACC saw.3s

(Don’t treat the case-marking here too seriously: it would have to be determined by language-specific rules.)

Or it might even use a completely different structure to English, say by using switch-reference and eliding some of the arguments:

[John-NOM ball-COM come-DS] saw-1s, [threw-DS] Kate-NOM saw-3s

(Where ‘DS’ means ‘different.subject’, indicating that this clause is subordinate to the following clause but has a different subject.)

And then of course there’s numerous other ways of doing it: the comitative could be at the beginning of the sentence or after the verb (assuming the language isn’t rigidly V-final, which you said it isn’t), or it might use a different conjunction (or none, as in the last example), or nominalised verbs might take their arguments in different ways, or sentence-like dependent clauses might be marked using something other than a complementiser, or be unmarked… the possibilities are endless!

So, my mind says this ought to be:

Code: Select all

pron.1st.SG.NOM J.ACC ball.SG.COM come.SG.PRES see.SG.PST, but K.NOM J.ACC pron.3rd.SG throw.SG.PRES see.SG.PST

This would be a fine alternative too. Explicitly indicating the clause structure (and condensing the gloss a little bit):

I John-ACC [ball-COM come.SG] saw.SG, but Kate-NOM John-ACC [he throw.SG] see.SG

Which shows that this is assuming a language where complement clauses have a sentence-like form, but the case-marking indicates that raising-to-object also takes place (or possibly subject control; I’m not entirely sure which one).

[hwhatting]

On the other hand, if it treated them as nominalisations, it might look more like:

I-NOM [John’s ball-COM come-ing]-ACC saw.1s, but Kate-NOM [John’s throw-ing]-ACC saw.3s

That's the kind of construction you'll find in Turkic languages.

[bradrn]

On the other hand, if it treated them as nominalisations, it might look more like:

I-NOM [John’s ball-COM come-ing]-ACC saw.1s, but Kate-NOM [John’s throw-ing]-ACC saw.3s

That's the kind of construction you'll find in Turkic languages.

Is it? I very strongly associate extensive use of nominalisations with Sino–Tibetan; Turkic to me suggests clause-chaining, though of course both families have both constructions. (I have a funny feeling Nortaneous might come and correct me, though.)