Word Order In Natlangs

[LinguoFranco]

So, word order is one of the first things conlangers learn about grammar and syntax. However, I think it's an area that is often understudied outside of correlations with word order and other parts of grammar.

For instance, one learns that some languages are pretty loose with their syntax despite having a preferred order they default to. Or they learn that SOV languages tend to prefer postpositions and cases, or that VSO languages are often head-marking and prefer prepositions. Simple stuff like that.

I don't think discussions about word order in conlanging really goes more in depth than that.

For my conlangs, I don't put too much thought into it. Usually, I just pick a word order as the default, and the decide whether it's strict or flexible, and then call it day.

I know that SOV languages often alternate to OSV under some circumstances, as with VSO to SVO. If I'm making an analytic language, I just say it's SVO (and maybe also topic-prominent) and leave it at that.

I'm wondering if there are any cool or underutilized things natlangs do with their word orders that could serve as inspiration for conlangers?

[Salmoneus]

Word order is an immense part of a language. For some languages, by far the largest part! It's not about finding one or two cool things, it's a whole... thing.

The three big things you're overlooking are:


- WHY is word order 'flexible'? It's probably not random, or purely aesthetic (outside of poetry). So why is there deviation from a language's default order? Which deviations occur, and when? Which deviations can co-occur, or MUST co-occur?


[I'm currently writing up word order for a Germanic language. I'm having to talk about, among other things, which elements have to go in P2 and in which order, fronting of phrases, or fronting of extracted elements of phrases (and why? and what can't be fronted/extracted?), shift of verbs to P2, leftward shift of adverbs, shift of elements to the end of the clause (and when? both semantic and prosodic motivations!), differences between main and subordinate clauses, etc etc]

- word order isn't just clause order. You have order of elements inside phrases as well - so Germanic noun phrases, for instance, have adjectives, possessors, prepositional phrases, relative clauses, comparative clauses and phrases, equative clauses and phrases, numerals, articles, possesive and deictic adjectives, quantifiers, and then you have most of those modifying most of the other ones, plus infinitive phrases popping up in several of them. And you also have ordering of different clauses within the sentence!

- word order in a clause needn't just be SOV. Indirect objects. Oblique arguments. Adverbs (of varying types and scopes). Various ways of sticking one clause inside another - raising, control, embedding. Verbal nouns, infinitives, gerunds. Conjunctions (of varying types). Etc.

[Creyeditor]

What Sal said. This is why linguists often talk about syntax, since word order is complicated and often interacts with other stuff. Here are some small tidbits that I found interesting:

• In some languages (e.g. Niuean, Tamil, Korean), case marked objects are more flexible in word order than unmarked objects. This is known as pseudo-noun incorpration. Word order interacts with morphology here.
• In some languages (e.g. Romance languages), object pronouns have different word order requirements than full-blown object nouns. Word order interacts with a lexical distinction here.
• In some languages (e.g. English), prosodically heavy (i.e. very long object noun phrases) have other ordering possibilities than lighter/shorter ones. Word order interacts with prosody here.
• In some languages (e.g. German, Yoruba, Latin) , certain words, e.g. German 'aber' but, can occur in a prosodically determined position instead of the word order expected purely on grounds of dependencies or constituent structures. Again, word order and prosody interact.
• In some languages (e.g. German, Greek) the order of possessor and possessed noun is determined by information structure to at least some extend. Noun phrase internal word order is determined by information structure.
• In some languages (e.g. German, English) the order of direct and indirect objects is determined by facts relating to information structure, animacy, pronounhood, etc. This is often known as scrambling. A lot of interaction here again.

[Salmoneus]

[*]In some languages (e.g. German, Yoruba, Latin) , certain words, e.g. German 'aber' but, can occur in a prosodically determined position instead of the word order expected purely on grounds of dependencies or constituent structures.

Wait, could you expand on that? [is this a relic of Wackernagel's position?]


One thing I didn't mention but that is important is how (or whether) a language allows elements to be extracted from their main clause entirely, and what apparatus gets placed around them when that happens. In formal English, an example of this is the creation of clefting constructions with dummy copula and pronoun in order to (amongst other things) create explicit topics - "it's the weather I'm worried about". In informal English, another example is simple bare extraction with a pronounced intonational marking: "This rain - I hate it!"


More generally, and the reason I didn't think of this the first time around, you should bear in mind that syntax (including word order) can be address from two different directions. One way is to think in terms of the ways phrases, clauses and sentences can be constructed (and then to explain when each construction might occu). But the other way is to think in terms of functions, and work out what constructions are used for each function (eg how do you do information structure, how do you do questions, how do you...). In theory these two approaches should yield the same complete description - but in practice I recommend thinking in both ways, as things that seem obvious or not worth mentioning in one approach can be foregrounded by the other.

[Creyeditor]

[*]In some languages (e.g. German, Yoruba, Latin) , certain words, e.g. German 'aber' but, can occur in a prosodically determined position instead of the word order expected purely on grounds of dependencies or constituent structures.

Wait, could you expand on that? [is this a relic of Wackernagel's position?]

I don't know if it's a relic. It's common in both formal and colloquial German, with only slight differences. All of the following is from Philipp Weisser's work. Consider the following sentence:

Maria will tanzen, aber ich will nach Hause gehen.
mary want dance but 1SG want to home go
'Mary wants to dance but I want to go home.

Now, 'aber' can also occur in several other positions, without a semantic difference.

Maria will tanzen, ich aber will nach Hause gehen.
mary want dance 1SG but want to home go
'Mary wants to dance but I want to go home.

Maria will tanzen, ich will aber nach Hause gehen.
mary want dance 1SG want but to home go
'Mary wants to dance but I want to go home.

In more complex sentences there are even more possible positions, e.g. between two objects.

Peter will tanzen, Maria will ihm aber den DJ vorstellen.
peter want dance mary want 3SG the DJ introduce
'Peter wants to dance but Mary wants to introduce the DJ to him.'

One could say this is a parataxis and 'aber' is actually some modal/discourse particle but this is not the position that these German particles usually occur in. There is also this cool sentence that shows that this 'aber' is sometimes required and not optional. (All of the above sentences work without 'aber' but the semantics are slightly different.)

Was hat sich Peter zum Geburtstag gewünscht, Maria ihm aber nicht gekauft?
what has himself peter to birthday wished mary him but not bought?
'‘What did Peter wish for for his birthday and Maria didn’t buy him?’

*Was hat sich Peter zum Geburtstag gewünscht, Maria ihm nicht gekauft?
what has himself peter to birthday wished but mary him not bought?

I know that Latin, Polish, and Yoruba do something similar, but I don't have the data at hand.

[Salmoneus]

Oh. I guess since that just looks like English that I wouldn't have thought of it as 'a thing'. I think this may be another instance where, being entirely ignorant of linguistics, I fail to understand how many (/most/all) languages are weird and ill-behaved by the standards of linguists... although I guess it's true that there are some languages with rigid adverbial locations.


[
"Maria wants to dance, though I want to go home"
"Maria wants to dance; I, though, want to go home"
"I don't want to dance; I want, though, to go home"
?"I don't want to go abroad; I want to, though, go home"
"Maria wants to dance; I want to home, though"

The commas are mostly for syntactic purposes, and don't necessarily imply an intonation or tempo distinction in most cases; likewise the comma/semicolon distinction might not have any audible significance and most speakers wouldn't know the 'correct' punctuation. The fourth example could be found in casual speech or in literature but not usually in normal writing and some may find it ungramamtical; it usually does require prosodic markers.

All that's happening here in English is that English has two syntactic classes - conjunctions and adverbs - which have different distributions. Adverbs can move quite freely through a clause, though this is seen as a form of extraction/interruption in many cases, hence the parenthetical commas; they most easily go after the first element after the verb, or at the end of the clause. This sort of adverb can't normally go at the front of the clause (though others can). Some words though, like "though", can zero-derive as either conjugations or adverbs (or, equivalently, are their own class with traits of both). That is, "though" can take the distribution of "but" OR that of "however".

Likewise, "Where did Peter say he wanted to go though he had never shown any signs of it before?" has a necessary "though". This does seem a difference from German, though, in that only the conjunctional use of "though" is allowed here; the adverbial use forces the two clauses apart, which means that the complex question fails.]

[Omzinesý]

I once considered a typology - from form to function - of word order as a thesis subject. Typology to that direction is very rare because there are very few formal constructions you can define between languages.

Word order codes many things.
- Information structure (somewhat in every language?)
- Definiteness (Chinese apparently)
- I found a paper about a langue where word order was the only clue of irrealis modality (Do we want to be Eurocentric, Germanic languages do that too.)
- question (German and Swedish)
- semantic roles of course
...

[Omzinesý]

Scandinavian languages change subject and object with the word order of the negation particle.

Anna äsker inte Maria.
A love NEG M
'Anna does not love Maria.'

Anna älsker Maria inte.
'Anna is not the one Maria loves.'

--

SOV word order is surprisingly common in Finnish subordinate clauses (at least those that follow the main clause). My introspection does not tell what is the difference in meanings between the normal SVO. Something very subtle. Maybe a starting change.

[eldin raigmore]

Anna äsker inte Maria.
A love NEG M
'Maria does not love Maria.'

On the third line I think one of your Marias needs to be an Anna! (Probably the first one!)

[Creyeditor]

Oh. I guess since that just looks like English that I wouldn't have thought of it as 'a thing'. I think this may be another instance where, being entirely ignorant of linguistics, I fail to understand how many (/most/all) languages are weird and ill-behaved by the standards of linguists... although I guess it's true that there are some languages with rigid adverbial locations.


[
"Maria wants to dance, though I want to go home"
"Maria wants to dance; I, though, want to go home"
"I don't want to dance; I want, though, to go home"
?"I don't want to go abroad; I want to, though, go home"
"Maria wants to dance; I want to home, though"

The commas are mostly for syntactic purposes, and don't necessarily imply an intonation or tempo distinction in most cases; likewise the comma/semicolon distinction might not have any audible significance and most speakers wouldn't know the 'correct' punctuation. The fourth example could be found in casual speech or in literature but not usually in normal writing and some may find it ungramamtical; it usually does require prosodic markers.

All that's happening here in English is that English has two syntactic classes - conjunctions and adverbs - which have different distributions. Adverbs can move quite freely through a clause, though this is seen as a form of extraction/interruption in many cases, hence the parenthetical commas; they most easily go after the first element after the verb, or at the end of the clause. This sort of adverb can't normally go at the front of the clause (though others can). Some words though, like "though", can zero-derive as either conjugations or adverbs (or, equivalently, are their own class with traits of both). That is, "though" can take the distribution of "but" OR that of "however".

Likewise, "Where did Peter say he wanted to go though he had never shown any signs of it before?" has a necessary "though". This does seem a difference from German, though, in that only the conjunctional use of "though" is allowed here; the adverbial use forces the two clauses apart, which means that the complex question fails.]

Two small things:
-German adverbials do not usually occur in all the positions that the floating conjunction occurs in, if I understand correctly what Weisser is trying to say.
-The last sentence is important because ot shows that the floating thing is a conjunction. Other adverbs do not license this kind of construction.

One could of course say that some conjunctions zero-derive a category adverb-B which is different from other adverbs in German. (This makes sense because different adverbs have different word order requirements in many languages (and in German) anyway (which is also an answer to the original question)). This would still be an interesting word order phenomenon.

[Sequor]

[*]In some languages (e.g. German, Yoruba, Latin) , certain words, e.g. German 'aber' but, can occur in a prosodically determined position instead of the word order expected purely on grounds of dependencies or constituent structures.

Wait, could you expand on that? [is this a relic of Wackernagel's position?]

I know that Latin, Polish, and Yoruba do something similar, but I don't have the data at hand.

For Latin, this surely has to do with Wackernagel's position. Incidentally, one day I was interested in what the famed Wackernagel's position actually entailed, since in Latin there is definitely a difference between words that attach to the first word quite simply (-que, -ve, -ne; they can even attach to a clause-initial short preposition: in-que urbe), and words that (must) attach to the first stressed word (enim, autem, the conjunction vero; you can't say *in enim urbe, it needs to be "in urbe enim"), besides the personal pronouns which can attach to the first stressed word but don't have to at all.

It turned out that Wackernagel's 1892 "paper" is a 100-page monograph covering all these and more (and 70% of the paper is dedicated to the phenomenon as found in Ancient Greek, the rest is Sanskrit and Latin). The uses of quidem '[noun] of course' and quoque 'too, also', which are found phonologically attached to a previous word (not at all necessarily the first stressed word, it could be anywhere in a clause) are also discussed. Wackernagel's position is about "second" position after another word in some vague but still useful sense.

Incidentally, something I find rather interesting is that the copula (or at least the present-tense 3rd-person forms est and sunt) may not count as phonologically salient enough and may be ignored for the Wackernagel's position in Latin. So you get things like:

Hoc enim est corpus meum
this for is my body
or:
Hoc est enim corpus meum
this is for my body
'For this is my body.' (famous line uttered by Jesus in the (translation of the) Gospel)

Duo sunt enim crimina...
two.NOM are for crimes.NOM...
'For two are the crimes [imputed against Clodia, of gold... and poison...] (Cicero, Pro Caelio 51)

Similarly, a sentence-initial non 'not' may be ignored...

Non excessisse enim Ponto Pharnacen audiebat
not exit.PRF.INF for Pontus.ABL Pharnaces.ACC heard.IMPF-3S
'For he didn't hear that Pharnaces had exited Pontus' (Aulus Hirtius, Bellum Alexandrinum 65)
(This could have also been "Non enim excessisse...")

[Salmoneus]

[*]In some languages (e.g. German, Yoruba, Latin) , certain words, e.g. German 'aber' but, can occur in a prosodically determined position instead of the word order expected purely on grounds of dependencies or constituent structures.

Wait, could you expand on that? [is this a relic of Wackernagel's position?]

I know that Latin, Polish, and Yoruba do something similar, but I don't have the data at hand.

For Latin, this surely has to do with Wackernagel's position. Incidentally, one day I was interested in what the famed Wackernagel's position actually entailed, since in Latin there is definitely a difference between words that attach to the first word quite simply (-que, -ve, -ne; they can even attach to a clause-initial short preposition: in-que urbe), and words that (must) attach to the first stressed word (enim, autem, the conjunction vero; you can't say *in enim urbe, it needs to be "in urbe enim"), besides the personal pronouns which can attach to the first stressed word but don't have to at all.

To hazard an amateur guess, this is because two different things are going on here. One is that some PIE particles specifically gravitate to second position; the other is that some particles (or even words) lack intrinsic stress and attempt to form phonological words with preceding (or following!) words. I suspect that the issue here is a two-stage process: there's the original "2nd position no matter what" particles that trigger tmesis (they can 'cut' a fronted phrase to maintain second position), and then there's other particles that became cliticised to stressed words and THEN gravitated to second position, while retaining the 'following a stressed word' rule.

For what it's worth, I seem to recall there are also words in Latin that appear to have preceded stressed words and attracted stress onto themselves - including of course the prepositional prefixes but also AFAIR some other words.

Incidentally, something I find rather interesting is that the copula (or at least the present-tense 3rd-person forms est and sunt) may not count as phonologically salient enough and may be ignored for the Wackernagel's position in Latin.

As I understand it, this is because the copula itself is often considered an unstressed clitic in IE languages.

Similarly, a sentence-initial non 'not' may be ignored...

Non excessisse enim Ponto Pharnacen audiebat
not exit.PRF.INF for Pontus.ABL Pharnaces.ACC heard.IMPF-3S
'For he didn't hear that Pharnaces had exited Pontus' (Aulus Hirtius, Bellum Alexandrinum 65)
(This could have also been "Non enim excessisse...")

This is definitely A Thing in IE languages. The negative particle didn't originally count for the V2 in Germanic, or the V1 rule in Celtic, and also tended to fuse with following common verbs - so "nill" in middle English (will we or nill we) an "nolo" in Latin.

----------

I don't suppose you've seen a clear list of P2-seeking or cliticising words in Latin, Greek or other IE language, have you?

And in particular, I've seen passing references to the clitic chains found in Anatolian, but not a concise description of what they contain.

[I also found Wackernagel's paper once upon a time, but IIRC got frustrated by the whole "I'll just assume you can read the greek alphabet and speak fluent greek" thing old linguists used to dp...]

[Creyeditor]

Today, I came across another interesting word order thing that natlangs do.

Many, if not most colloquial varieties of German have a periphrastic progressive construction that looks something like the following sentence. An auxiliary is followed by the preposition am followed by an infinitive.

Ich bin am telefonieren.
1SG am at talking.on.the.phone.INF
'I am on the phone right now.'

So far so good. But what to do with objects? In some varieties of colloquial German the object position is heavily restricted here. Basicall, only single nouns can occur, like in the following sentence. These object nouns occur between the preposition and the verb.

Ich bin am Nudeln essen.
1SG am at pasta eat.INF
'I am eating pasta.'

Other varieties of colloquial German are more flexible when it comes to object noun phrases. More complex noun phrases, however, usually trigger a slightly different word order. They occur before the preposition.

Die waren da die eine Brücke am reparieren.
they were there that one bridge at repair.INF
'At that point, they were repairing that one bridge.'

I think both patterns are interesting. The first group of varieties restricts what can occur in the object position in certain constructions, the other group of varieties have slightly different word orders depending on properties of the object.

Sal is probably going to tell me that the same (or something similar) happens in English, but I am not a native English speaker, so I would be delighted to hear about it Especially the varieties of English that use forms like 'The train is a'coming.'

[Salmoneus]

Well, since you mention it!


So, the "a-Xing" construction is historically a thing in English, yes, although it's rarely used today (I imagine some dialects still use it - I would assume in the American South and the English North and West, and maybe Ireland). It may actually have two convergent sources in English - a more ancestral Germanic source from prefixed an-, and a more recent source in 'at' that's calqued from Irish use of verbal nouns.


I think that what's "happening" in your examples (i.e. a story I would tell to highlight analogous processes in other languages) is that cross-linguistically objects of verbal nouns/gerunds/etc are often incorporated into the noun, or locked into constructions that aren't officially incorporation but that are a bit like it (eg there's a very strong tendency to not put argument marking on such nouns, as though the verb were ergative, which could alternatively be explained by saying that the 'noun' is in a nominal compound phrase, not in a verb phrase where argument marking could be found). And there's usually restrictions on which nouns can easily be incorporated - i.e. only simple nouns, not entire nominal phrases.

So, I think that some German dialects maybe do this: they incorporate object nouns into the gerunds, even though they don't spell them as one word (and though they may not phonologically be one word). This can only happen with simple nouns due to the incorporation rules. Therefore, gerunds can't have non-simple objects.

But in other German dialects there has been a desire to be able to have gerunds with non-simple objects... so there are. But these non-simple objects still can't be incorporated, so a different, more directly verb-like structure is used instead.

Why does that have anything to do with the position of the preposition? Well, that reinforces my interpretation. The preposition forms a unit with the gerund - we could call it a gerundal phrase, or we could even say the preposition is actually part of the word (it's not like Germanic prepositions don't have an ambiguous independence from the verb in many other contexts too! it's also of course ambiguous whether this IS a preposition, rather than a derivational particle/prefix that creates this kind of gerund!). Either way, the preposition directly precedes the gerund. In the simple-object scenario, the object is incorporated, and therefore part of the gerund, so the preposition directly precedes it. But in the complex-object scenario, in dialects where that's legal, the object is not incorporated, so isn't part of the gerund, so the preposition can't precede it, because then it wouldn't be directly preceding the gerund, so it has to follow the complex object phrase instead.

[conlang idea: complex noun phrases take accusative marking, because it's a verb-like structure, while simple noun arguments take absolutive (zero) marking because they're really incorporated...]

And how does this relate to English? Well, people's first answer will be that it doesn't apply, because objects don't precede the gerund, so the situation doesn't arise. Except, of course, it does! Except in English the simple object is admitted to have been incorporated!

Hence, some imaginary but probably approximately accurate historical and/or regional dialect:
? That dog's a-water-drinkin' again I see!
vs
? That dog's a-drinking o' the neighbour's water again I see!

Certainly in standard English the equivalent distinction is made with adjectives (/participles/relatives/whatever):
the fish-eating cat is black or the cat eating fish is black
but only
the cat eating the neighbour's fish is black

The simple noun can be incorporated and hence precede the adjective that precedes the noun, but the complex noun phrase cannot be, and so an entirely different word order is required. [this incorporation arguably comes with a small amount of prosodic marking - you could disambiguate "the cat-eating fish" from "the cat eating fish" by intonation and tempo if you needed to - but I'm not sure how much of that distinction is actually present in non-disambiguatory allegro speech]


-------

There's also an interesting parallel here with another "preposition"-verb construction in English - phrasal verbs!

I ripped the bandage off or I ripped off the bandage
but
? I ripped the orange bandage that the Prime Minister's cousin had sold me the weekend before off
is theoretically grammatical at least in writing, but in practice, particularly in speech, would be greatly disprefered in favour of
I ripped off the orange bandage that the Prime Minister's cousin had sold me the weekend before
[and when the former is occasionally used, the prosody is different from the case with the simple object phrase, with 'off' becoming stressed]

Here, though, it's not a binary distinction: the longer and more complex the object phrase, the less likely it is to precede the adverb. This may have something to do with a looser form of "incorporation" (not as one word but as a phrase of some sort) or it may simply be a practical constraint about distance between elements of a lexical unit. Such concerns may also be a factor in your gerundal example. Such concerns are famously NOT a factor in German separable verbs, but it's possible German simply has normalised separable verb syntax sufficiently as its own thing that it can sidestep such concerns, while they remain active in the case of gerunds.

-----------


Anyway, I hope we're demonstrating to the OP not only some specific word order issues, but also entire types of issues. Issues like "which things are independent words isn't always clear" (and syntactic vs phonological words may not be the same), or "what part of speech a particle belongs to may not be clear", or perhaps "prosodic differences may accompany syntactic differences - but are they distinctive in their own right, and are they always present or only in disambiguations?"

[Creyeditor]

I like how looking at very specific phenomena in a handful of languages already touches many larger areas that can serve as an inspiration for conlang syntax. Also, I definitely agree with the assessment that the German data are related to (pseudo-)noun incorporation. IIRC, the standard explanation involves Object-Verb-nmlz compounds and some kind backformation but I don't think the two are incompatible. And thanks for all the interesting English data of course.

[LinguoFranco]

So, I have been reading up on the word order of some languages like Selk'nam, Päri and Macushi. They are analyzed as being OVS, but according to Wikipedia, they are actually absolutive-verb-ergative.

What are your thoughts on this, and how do such a thing arise?

What is the ergative split based off of in these languages?

[Creyeditor]

I don't know about Päri, but many Nilotic languages have word order patter that are similar in their complexity to V2 patterns in Germanic languages and similar to voice systems in Philippine lamguages. Depending on the form of the verb, different word orders are possible.