Adpositions and Relative Clauses in natlangs

[dhok]

This is not a harangue about "to which..." vs. "which... to", though it is looking at it descriptively.

Basically, in all languages save English I know of (there I go again!) that have adpositions (A) and relative pronouns (R), we've got a system where the preposition and the pronoun are right next to each other, and the clause never comes between them. What I don't see outside of English is R-Clause-A or A-Clause-R; that is, are there any non-English natlangs that do this as a matter of routine?

[TomHChappell]

Does
http://wals.info/feature/combined/85/90
help answer your question?

[dhok]

I feel like it should, but I don't get the exact mechanics of what it's surveying.

[Mecislau]

This is not a harangue about "to which..." vs. "which... to", though it is looking at it descriptively.

Basically, in all languages save English I know of (there I go again!) that have adpositions (A) and relative pronouns (R), we've got a system where the preposition and the pronoun are right next to each other, and the clause never comes between them. What I don't see outside of English is R-Clause-A or A-Clause-R; that is, are there any non-English natlangs that do this as a matter of routine?

The very first place I thought to look:

This construction is widely found in Germanic languages, including English[1] and the Scandinavian languages;[2][3] whether or not German and Dutch exhibit legitimate preposition stranding is debatable. Preposition stranding is also found in languages outside the Germanic family, such as Vata and Gbadi, two languages in the Niger-Congo family, and certain dialects of French spoken in North America.

[linguoboy]

What's your take on resumptive pronouns? It's only in very formal Irish that I see constructions like "Sin í an bhean lenar labhair mé" ("That is the woman with whom I spoke"). What most people write and say these days is "Sin í an bhean ar labhair mé léi" ("That is the woman that I spoke with her"). [Irish has personal forms of most prepositions, so you can't strand them the same way you can in English.]

Modern colloquial Welsh doesn't even seem to have the first option. As far as I know, it's always "Dyma ddynes siarades i â hi."

[Mecislau]

What's your take on resumptive pronouns? It's only in very formal Irish that I see constructions like "Sin í an bhean lenar labhair mé" ("That is the woman with whom I spoke"). What most people write and say these days is "Sin í an bhean ar labhair mé léi" ("That is the woman that I spoke with her"). [Irish has personal forms of most prepositions, so you can't strand them the same way you can in English.]

Modern colloquial Welsh doesn't even seem to have the first option. As far as I know, it's always "Dyma ddynes siarades i â hi."

Modern Hebrew does the same thing:

זוהי האישה שדיברתי איתה
Zohi ha'iša šedibarti 'itah
This_is.fem def-woman rel-speak.1sg.past with-3sgfem
"This is the woman that I spoke with her"

[Yng]

Modern colloquial Welsh doesn't even seem to have the first option. As far as I know, it's always "Dyma ddynes siarades i â hi."

Some people accept preposition stranding, but it's eww. I can't actually think (other than maybe with interrogatives) of any cases, colloquial or otherwise, which don't work like that.

[schwhatever]

It seems like there could be a potential shift (à la agglutinating inflecting isolating agglutinating) here:

Stage 1: alternative component order with A-R/R-A sequence/forms/whatever
"That's the woman to whom I gave it."

This system requires a more complex relative conjunction system and usually some flexibility with word order. In short, it requires that a language avoid general isolating morphology, since the word order needs to be able to be modified.

Stage 2: simplified conjunctions with resumptive pronouns
"That's the woman that I gave it to her."
Let's say the language's more complicated morphology breaks down (as they are often wont to do). The relative conjunctives themselves might be part of that breakdown, but in either case, word order starts mattering more, and the second clause is more likely to "insist" on normal word order, rather than a deviation from it.

Stage 3: pro-drop solutions with R-clause-A//A-clause-R sequences
"That's the women that I gave it to."

Resumptive pronouns were a fine solution, but they're a redundancy, and if the language flirts with becoming more pro-drop (like English sometimes does), it could easily decide to stop including the second pronoun, but otherwise leave the construction as is.

There's some clear gray areas though - English shows that a language can easily have multiple solutions coexist, and furthermore seems to have progressed directly from stage 1 to stage 3. Maybe English interpreted replacement of (for example) "to whom" with "that/who ... to X" as reduplicating the pronoun (which we usually don't do) and hence did everything but that?

Likewise, is there any actual evidence of this being a process? Other than highly tentatively Welsh? (Among other things - it seems like it could just be English influence).

[Yng]

The only reason that preposition stranding is becoming grammatical in Welsh is because of English-Welsh bilingualism with higher competence in English, I think.

[TomHChappell]

I feel like it should, but I don't get the exact mechanics of what it's surveying.

Well, what exactly were you asking for?

Wals.info/features/combined/85/90 shows that out of the 624 languages in their sample database for which they recorded both features 85 (order of adposition and noun) and 90 (order of relative clause and noun), 311 of them (including English) have prepositions and put the RC after the noun it modifies (first column, second row), and another 100 have postpositions and put the RC before the noun it modifies (second column, first row).

If the "noun" is actually a relative pronoun -- a pronoun that acts also as a relativizer -- then in these languages one would logically expect most of them to usually have an adposition next to a relative pronoun, in Adp-RelPron-RC or RC-RelPron-Adp constructions.

But I have begun to doubt, given the posts here since my previous post -- especially given your posts -- whether you really wanted to ask about relative clauses and relativizers. I wonder whether you may have wanted instead to ask about complement clauses and complementizers.

If that's the case, look at WALS.info's features 122 through 128, especially 124 "'Want' Complement Subjects" and 128 "Utterance Complement Clauses". I am sure you'll need to read the Chapter Text for feature 128 to understand it; at least I did. Feature 124 may be more self-explanatory.

OTOH if you really were intending to be asking about relative clauses, features 60, 90, 96, 122, and 123 are relevant.

Read http://wals.info/supplement/intro122-123 and http://wals.info/feature/description/122, and maybe also http://wals.info/feature/description/123.

Features 122 and 123 are talking about restrictive relative clauses only, not attributive relative clauses. That is, "...a clause narrowing the potential reference of a referring expression by restricting the reference to those referents of which a particular proposition is true".

In Feature 122 the first strategy, "Relative Pronouns", is described thusly:

The first strategy has come to be called the relative pronoun strategy: the position relativized is indicated inside the relative clause by means of a clause-initial pronominal element, and this pronominal element is case-marked (by case or by an adposition) to indicate the role of the head noun within the relative clause.

That would impy that all 12 of the languages (Acoma, Bulgarian, English, Finnish, French, Georgian, German, Greek (Modern), Hungarian, Italian, Latvian, and Russian) shown on their map as having the Relative Pronoun strategy for Relativizing on Subjects, would have adpositions next to relative pronouns at the beginnings of many of their (at least their restrictive) relative clauses, without the rest of the RC intervening between the adposition and the RC.

Combining Features 85 and 122 shows that four of these languages (Acoma, Finnish, Georgian, and Hungarian) have postpositions while eight of them (Bulgarian, English, French, German, Greek (Modern), Italian, Latvian, and Russian) have prepositions. I don't know if that's relevant to what you were asking?

Looking at 123 (Relativization on Obliques), we see that 13 of their sample database's languages (Acoma, Bulgarian, English, Finnish, French, Georgian, German, Greek (Modern), Hungarian, Italian, Latvian, Russian, and Spanish) are recorded as using the Relative Pronoun strategy to relativize on "obliques". But "obliques" may not mean "oblique arguments other than Subjects, Direct Objects, or Indirect Objects" in this WALS.info feature; it may just mean "arguments other than Subjects" here.

Except for Spanish, the languages they've recorded as using the Relative Pronoun strategy for Subjects are just the same as the languages recorded as using RelProns for Obliques. (Spanish is recorded as using Gapping for relativizing on Subjects, but RelProns for relativizing on Obliques. No reference has been given to back this up; do any Spanish-speakers here know whether that's accurate?)
Combining 85 and 123 shows that the same 4 languages have postpositions, and the same 8 languages plus Spanish have prepositions.

A construction is considered a relative clause for the purposes of this map if it is a clause which, either alone or in combination with a noun, denotes something and if the thing denoted has a semantic role within the relative clause. If there is a noun inside or outside the relative clause that denotes the thing also denoted by the clause, that noun will be referred to as the head of the relative clause. Headless relative clauses (like English "what I bought at the store" ) are not relevant to this map.

Combining features 90 (order of RC and Noun) and 122 shows that of the 10 languages recorded as having a RelPron strategy for Subjects and also as having a recorded value for Feature 90, nine (including English) use the order Noun-RC and one (Hungarian) uses a Mixed order of RC and Noun.

Combining features 90 (order of RC and Noun) and 123 shows pretty much the same thing, with Spanish added to the majority.

Does any of that help?
How much of it helps? (Most of it, I hope.)

--------------------------------------------------------------------------------------------------------------------------------------------------------------

As for resumptive pronouns, languages with resumptive pronouns are coded in Features 122 and 123 as having a "Pronoun-Retention Strategy". Five languages (Babungo, Baka (in Cameroon), Kayah Li (Eastern), Ngemba, and Yoruba) are recorded as using resumptive pronouns for relativizing on subjects, and twenty languages (Anywa, Arabic (Egyptian), Babungo, Bagirmi, Baka (in Cameroon), Bole, Ewe (Anglo), Gaelic (Scots), Guaraní, Hausa, Hebrew (Modern), Irish, Kayah Li (Eastern), Krongo, Lango, Lingala, Luo, Miya, Paamese, and Persian) are recorded as using resumptive pronouns for relativiziong on obliques.

All three of the languages recorded as using resumptive pronouns for relativizing subjects of RCs and recorded as having a dominant order of adposition and noun are recorded as prepositional in WALS.info.

Fourteen of the fifteen languages recorded as using resumptive pronouns for relativizing obliques of RCs and recorded as having a dominant order of adposition and noun are recorded as prepositional in WALS.info.


--------------------------------------------------------------------------------------------------------------------------------------------------------------

As for adposition-stranding, I'm pretty sure WALS.info doesn't say anything about that, at least not yet.

[Ser]

Except for Spanish, the languages they've recorded as using the Relative Pronoun strategy for Subjects are just the same as the languages recorded as using RelProns for Obliques. (Spanish is recorded as using Gapping for relativizing on Subjects, but RelProns for relativizing on Obliques. No reference has been given to back this up; do any Spanish-speakers here know whether that's accurate?)

It is, but then again, according to how they've defined those terms. "Gapping" here includes having a relative pronoun that doesn't change for case (look at the example from the Maybrat language in (8) in the respective article).

That's what was pointed out in Viktor77's "French to Spanish relative pronouns" thread. In Spanish, "que" works as the relative pronoun regardless of whether the antecedent has a role of the subject or direct object in the relative clause. Then, what differentiates these roles is the lack of a gap when it has the role of the subject inside the relative clause.

Using the same sentence as in (8):

la mujer que come malanga
the woman REL 3.SG.eat taro

[Bedelato]

It seems like there could be a potential shift (à la agglutinating inflecting isolating agglutinating) here:

Stage 1: alternative component order with A-R/R-A sequence/forms/whatever
"That's the woman to whom I gave it."

This system requires a more complex relative conjunction system and usually some flexibility with word order. In short, it requires that a language avoid general isolating morphology, since the word order needs to be able to be modified.

Stage 2: simplified conjunctions with resumptive pronouns
"That's the woman that I gave it to her."
Let's say the language's more complicated morphology breaks down (as they are often wont to do). The relative conjunctives themselves might be part of that breakdown, but in either case, word order starts mattering more, and the second clause is more likely to "insist" on normal word order, rather than a deviation from it.

Stage 3: pro-drop solutions with R-clause-A//A-clause-R sequences
"That's the women that I gave it to."

Apparently nominal number inflection also broke down…

But anyway, Stage 2 actually pops up sometimes IMI:
"There's this person who I don't know his name".

I believe I mentioned something like this before, but it's relevant right now and I just couldn't resist

[schwhatever]

Nominal number? How so?

The only reason that preposition stranding is becoming grammatical in Welsh is because of English-Welsh bilingualism with higher competence in English, I think.

Yeah, I don't feel comfortable considering it a process until we have another (more geographically distant) example. I'm worried that it'll be hard to rule out English influence if the change is too recent anywhere though.

Somewhat related, was the Modern Hebrew system with resumptive pronouns a response to the breakdown of more specific relative conjunctions? I vaguely remember Hebrew doing all sorts of weird stuff with pronouns (fusing them to prepositions, for instance) so I don't really know what conclusion can be drawn there.

[Mecislau]

Somewhat related, was the Modern Hebrew system with resumptive pronouns a response to the breakdown of more specific relative conjunctions? I vaguely remember Hebrew doing all sorts of weird stuff with pronouns (fusing them to prepositions, for instance) so I don't really know what conclusion can be drawn there.

No. Hebrew doesn't appear to have even had any true relative conjunctions, just subordinating conjugations like 'ašer (modern še-). Since these aren't actually pronouns, but rather just a means of linking together two clauses, all pronouns are left in place in the subordinate clause:

"That is the woman. I spoke with her" > "That is the woman 'ašer I spoke with her"


(And yes, Hebrew does merge prepositions and pronouns, but that phenomenon is very, very old; it's in the other Semitic languages as well)


The only exception to leaving in pronouns in the subordinated clause is when the noun in the main clause is the direct object of the subclause (e.g., "That's the book which I read [it]". I'm not sure, but this might be a consequence of how direct objects used to be marked on the verb (so you'd have "I-read-it"), but no longer are in Modern Hebrew. Or it could be completely unrelated and just a normal simplification.

[Ser]

Resumptive/Anaphoric pronouns (does anybody know whether one of these terms is "better"?) can be found in Written Arabic too, which somewhat suggests their use could be quite old in Semitic. The subordinators don't seem to be cognates at all though.

هذه هي المرأة. تحدثت إليها.‏
/haːðihi hija l-marʔatu/ /taħaddaθtu ʔilaj-haː/
This is the-woman. I.spoke with-her.

هذه هي المرأة التي تحدثت إليها.‏
/haːðihi hija l-marʔatu llatiː taħaddaθtu ʔilaj-haː/
This is the-woman REL I.spoke with-her.

[Yng]

Welsh doesn't solely make use of resumptive pronouns - only with the subordinator y (or at least in the cases where that subordinator was historically used), which is used to extract all arguments that aren't subjects or objects. This includes the objects of verbnouns, which syntactically are treated as genitives (in fact, what you might call extracted genitives which in English have 'whose' and the extracted arguments of verbnouns are syntactically identical, at least historically):

Y dynes y siaradais i â hi
the woman SUB speak_1ps.PRET I with her
The woman I spoke with (her)

Y dynes y gwelais i ei gwr
the woman SUB see_1ps.PRET I her.POSS husband
The woman whose husband I saw (that I

Y dynes y gwnes i ei saethu
the woman SUB do_1ps.PRET I her.POSS shoot_VN
The woman who I shot (that I did her shooting - 'gwnes' is a form of 'gwneud', which acts here as an auxiliary for the preterite with identical meaning to the -ais suffix seen on 'gwel-' above)

But:

Y dynes a welodd i
the woman REL see_PRET.3ps me
the woman that saw me (note the mutation on 'gwelodd')

Y dynes a welais i
the woman REL see_PRET.1ps I
the woman that I saw

[linguoboy]

Yeah, I don't feel comfortable considering it a process until we have another (more geographically distant) example. I'm worried that it'll be hard to rule out English influence if the change is too recent anywhere though.

Not just English, but other Germanic languages as well. You have examples of stranding in both Alsatian French and Cajun French.

[schwhatever]

On that note, here's the only paper I could find that even briefly talks about non-European languages with this phenomenon. What it looks like is that Gbadi, one of the two Niger-Congo languages mentioned in the wikipedia particle, tends to produce subordinate clauses that associate (it's not clear to me if we're talking about morphologically, but at least syntactically) the position and the clause's verb - which is intriguingly similar to Germanic languages but quite sufficiently removed from them, that I would assume it's an independent innovation. That fits nicely into the point of the paper, which is that it's about (re?)interpreting the position as (for all intents and purposes) an adverb, rather than a position (which theoretically has a nominal head). It's not about pro-drop or anything.

Somewhat tangentially, the study talks about "swiping" - the phenomenon in phrases like "He was talking about cheese, but I don't know who with" - which raises some interesting questions about English. I can't speak for other dialects, but I find mine does allow resumptive pronouns in situations were "swiping" doesn't work, namely which. For example, IMD:

*Something crazy happened yesterday in some campus class, but I don't know which in
*Something crazy happened yesterday in some campus class, but I don't know in which
?Something crazy happened yesterday in some campus class, but I don't know in which one.
?Something crazy happened yesterday in some campus class, but I don't know which one in.

I think the fourth one works IMD, but I'm not sure. I'm more skeptical about the third, but again, not clear enough. I have a sinus infection, so I can't think clearly. Any thoughts?

[Cedh]

That remark about Gbadi and reanalysis of adpositions as adverbs sounds interesting.

wikipedia particle

[schwhatever]

AAAAH. Fracking sinus infection. Ok, I'm going to go sleep now.

[TomHChappell]

It is, but then again, according to how they've defined those terms. "Gapping" here includes
....
That's what was pointed out in Viktor77's "French to Spanish relative pronouns" thread. In Spanish, "que"
....

Thanks.