Linguistic Miscellany Thread

[akam chinjir]

I had assumed that differential subject marking was a general category of phenomena in which the subject can be marked in multiple different ways, encompassing both optional ergative marking and the presence of multiple ergative markers. Is this definition of DSM incorrect?

To be honest I'm not sure how broadly people use the term. I mostly associate it with DOM and the same kinds of alternations you get there, and maybe that's a mistake. Though I think the examples you give are very different in terms of their importance. Things like NP splits and different ergative markers on different nouns/pronouns seems likely to be just morphology, without much significance in synchronic grammar. On the other hand, the sort of pragmatically-conditioned case-marking you apparently get in Tibeto-Burman languages does seem really interesting. Maybe when this stage of the reconstruction relay is past I can have a go at learning more about it

[bradrn]

I had assumed that differential subject marking was a general category of phenomena in which the subject can be marked in multiple different ways, encompassing both optional ergative marking and the presence of multiple ergative markers. Is this definition of DSM incorrect?

To be honest I'm not sure how broadly people use the term. I mostly associate it with DOM and the same kinds of alternations you get there, and maybe that's a mistake. Though I think the examples you give are very different in terms of their importance. Things like NP splits and different ergative markers on different nouns/pronouns seems likely to be just morphology, without much significance in synchronic grammar. […]

I wasn’t talking about those things — they’re just ordinary split ergativity. I was talking more about the phenomena I listed in the last section of this post, where ergative can be marked in two (or more) ways, each having different semantics or pragmatics (my list there includes agentivity, focus, social expectation and personal relevance as possible meanings, along with differential marking based on O).

[Kuchigakatai]

One thing I've never heard about is differential modification with relative clauses. Standard Arabic has it, where definite nouns must be followed by the subordinator of relative clauses (which is an adjective agreeing with the head noun, funnily enough), but indefinite nouns take the relative clause directly.

الجبل الذي رأيته
/al-dʒabal al:aði: raʔaitu=hu/
the-mountain SUB_REL.MASC.SG saw.1SG=3SG
'the mountain I saw'

جبل رأيته
/dʒabal raʔaitu=hu/
mountain saw.1SG=3SG
'a mountain I saw'

[akam chinjir]

the last section of this post, where ergative can be marked in two (or more) ways, each having different semantics or pragmatics (my list there includes agentivity, focus, social expectation and personal relevance as possible meanings, along with differential marking based on O).

Ah, I'll definitely have to find time to look more into those things.

[chris_notts]

Well, my conlang doesn’t have verbal agreement. I suppose I’ll have to build some other sort of reduced reference form in then.

(Oh, and a question: how do you tell the difference between verbal agreement and a clitic pronoun? Surely they would be very similar.)

Normally the difference is that:

(a) clitics are less phonologically integrated with their hosts than affixes
(b) clitics normally alternate (i.e you don't see both clitic + independent argument the way you do with agreement)

Alternation would mean:

I saw=him
I saw the man
*I saw=him the man

Quick question: what do you mean by a ‘covert subject’?

I meant that, even if morphology or word order doesn't mark a "subject" case or role, syntactic behaviour may suggest that a subject role exists. Most ergative languages show either neutral or accusative alignment in their syntactic control constructions.

I’m not sure I understand what you’re saying here. Are you saying that topic-prominent languages less often have argument-rearranging constructions like passive and antipassive, but are rich in valency-changing constructions like causative and applicative? Or are you saying that topic-prominent languages can still have all these constructions, but don’t usually use them purely for argument rearrangement in the way that most non-topic-prominent languages do?

I'm saying that, if a language does not associate topicality with a GR, there's no reason to use devices like passives to promote topical arguments to that GR. The only reason to use them is either when they have other semantic effects (middle), or when they completely eliminate an unnecessary argument (de-transitivising passive) or add an argument (causative).

Also, once you bring up control: I always thought that in a topic-prominent language, the topic could be used as a pivot for control — that is, when you coordinate two clauses, it is the topics that get equated (rather than, say, S/A being equated). (So for example in imaginary topic-prominent English, ‘the man(topic) saw the woman and ran’ would have the man running, but ‘the woman(topic) the man saw and ran’ would have the woman running.) Is this attested? And if so, how common is it?

Yes, this is kind-of what happens, except in most languages I'm aware of with overt topic marking, the rules are not as strict as in a language like English where only strict subject identity allows zero anaphora. In most of these languages, pragmatics also play a role in identifying the controller, because there may be multiple competing highly topic referents in the discourse, so the controller is less strict and less syntactically restricted.

To make sure I understand: the first example here is informational focus while the second is contrastive focus, right?

Yes

A clarification here: I assumed earlier that each of these patterns was present over a whole language (i.e. much like an ergative alignment may be found on all clauses, similarly a language might always consistently mark focus rather than topic and so be a language with argument focus), but you seem to imply here that these patterns are properties of individual clauses and sentences rather than a property of a language as a whole. Which interpretation is correct?

Every language will have some way of distinguishing between cases like these, although some constructions may be ambiguous. Take English, for example. The most common strategy in spoken English is intonation, because English has rigid word order, lacks focus particles, and isn't a huge fan of clefts. But intonation has some ambiguity:

1. The man ate a pizza <-- normal intonation = topic - comment, subject = topic
2. The man ate a pizza <-- emphasis on subject can mark either subject contrastive focus or sentence focus
3. The man ate a pizza <-- emphasis on the object can only mark object contrastive focus

Clefts are more marked, but are unambiguously contrastive focus constructions so can disambiguate (2):

It was the man who ate the pizza <-- can only be subject contrastive focus

In other languages, word order may do what intonation does in English. For example, in Spanish, subject-verb inversion (placing the subject after the verb) is associated with focal subjects, and I believe (trying it out in my head here) shows the same ambiguity between subject contrastive focus and sentence focus that you get with intonation in English.

In Basque, as in many other verb final languages, the immediately pre-verbal position is favoured for focal arguments, new topics go first, and anything which is backgrounded can be shifted after the verb to the end of the clause.

[zompist]

And a question: how do you tell the difference between verbal agreement and a clitic pronoun? Surely they would be very similar.)

Normally the difference is that:

(a) clitics are less phonologically integrated with their hosts than affixes
(b) clitics normally alternate (i.e you don't see both clitic + independent argument the way you do with agreement)

Alternation would mean:

I saw=him
I saw the man
*I saw=him the man

I'm probably not understanding you, because I can immediately think of two languages where you say just this:

Colloquial French: Je l'ai vu, l'homme. Or even Je l'ai vu, l'homme, moi. You can omit either the l' or l'homme.

Tok Pisin: Em i lukim mi, "He looks at me". ("Lukim" derives from "look 'im" but has been reinterpreted as a marker of transitivity. But at some stage in the language, this presumably precisely paralleled "I saw him the man.")

[bradrn]

the last section of this post, where ergative can be marked in two (or more) ways, each having different semantics or pragmatics (my list there includes agentivity, focus, social expectation and personal relevance as possible meanings, along with differential marking based on O).

Ah, I'll definitely have to find time to look more into those things.

If you want some resources, I mainly relied on William B. McGregor’s work for that section, particularly Typology of Ergativity and Optional ergative case marking systems in a typological-semiotic perspective. I think I may have used the Oxford Handbook of Ergativity as well, but I believe you already read that one (you recommended it to me, I think).

Well, my conlang doesn’t have verbal agreement. I suppose I’ll have to build some other sort of reduced reference form in then.

(Oh, and a question: how do you tell the difference between verbal agreement and a clitic pronoun? Surely they would be very similar.)

Normally the difference is that:

(a) clitics are less phonologically integrated with their hosts than affixes
(b) clitics normally alternate (i.e you don't see both clitic + independent argument the way you do with agreement)

Alternation would mean:

I saw=him
I saw the man
*I saw=him the man

Oh, of course! I do remember now that clitics alternate with the independent argument but agreement affixes don’t. (Actually, I discussed that previously here, so I really should have remembered this.)

Quick question: what do you mean by a ‘covert subject’?

I meant that, even if morphology or word order doesn't mark a "subject" case or role, syntactic behaviour may suggest that a subject role exists. Most ergative languages show either neutral or accusative alignment in their syntactic control constructions.

I know what you’re talking about, but I haven’t ever heard it called a ‘covert subject’; I’ve always known it as ‘syntactic accusativity’ or ‘shallow ergativity’.

Also, once you bring up control: I always thought that in a topic-prominent language, the topic could be used as a pivot for control — that is, when you coordinate two clauses, it is the topics that get equated (rather than, say, S/A being equated). (So for example in imaginary topic-prominent English, ‘the man(topic) saw the woman and ran’ would have the man running, but ‘the woman(topic) the man saw and ran’ would have the woman running.) Is this attested? And if so, how common is it?

Yes, this is kind-of what happens, except in most languages I'm aware of with overt topic marking, the rules are not as strict as in a language like English where only strict subject identity allows zero anaphora. In most of these languages, pragmatics also play a role in identifying the controller, because there may be multiple competing highly topic referents in the discourse, so the controller is less strict and less syntactically restricted.

I’m not sure I understand what you’re saying here… how can there be multiple topics?

And a question: how do you tell the difference between verbal agreement and a clitic pronoun? Surely they would be very similar.)

Normally the difference is that:

(a) clitics are less phonologically integrated with their hosts than affixes
(b) clitics normally alternate (i.e you don't see both clitic + independent argument the way you do with agreement)

Alternation would mean:

I saw=him
I saw the man
*I saw=him the man

I'm probably not understanding you, because I can immediately think of two languages where you say just this:

Colloquial French: Je l'ai vu, l'homme. Or even Je l'ai vu, l'homme, moi. You can omit either the l' or l'homme.

Tok Pisin: Em i lukim mi, "He looks at me". ("Lukim" derives from "look 'im" but has been reinterpreted as a marker of transitivity. But at some stage in the language, this presumably precisely paralleled "I saw him the man.")

In that case, wouldn’t lʼ- and -im be (optional) agreement affixes rather than pronominal clitics? chris_notts’s point that I saw=him the man is ungrammatical only applies for clitics.

[akam chinjir]

And a question: how do you tell the difference between verbal agreement and a clitic pronoun? Surely they would be very similar.)

Normally the difference is that:

(a) clitics are less phonologically integrated with their hosts than affixes
(b) clitics normally alternate (i.e you don't see both clitic + independent argument the way you do with agreement)

Alternation would mean:

I saw=him
I saw the man
*I saw=him the man

I'm probably not understanding you, because I can immediately think of two languages where you say just this:

Colloquial French: Je l'ai vu, l'homme. Or even Je l'ai vu, l'homme, moi. You can omit either the l' or l'homme.

Tok Pisin: Em i lukim mi, "He looks at me". ("Lukim" derives from "look 'im" but has been reinterpreted as a marker of transitivity. But at some stage in the language, this presumably precisely paralleled "I saw him the man.")

The French case is a bit tricky, though: you can't do it without a bit of pause, right? That's sometimes called clitic right dislocation, and distiguished from clitic doubling, which is when you get the clitic even when there's an overt NP argument fully integrated in the same clause. But clitic doubling is also reasonably common, I think. Wikipedia has some examples from Spanish, like Le di un regalo a mi madre. Exactly when something is clitic doubling and when it's agreement can be a pretty subtle issue, as bradrn says they can look very similar.

[chris_notts]

Yes, maybe saying that clitics are never doubled is an exaggeration. In reality there's cline of obligatoriness and phonological integration:

Most clitic like ------------------------------------------------------------------------------- Most affix like
Not doubled ---------------------------------------------------------------------------------- Obligatory regardless of presence of coreferential noun phrase
Minimal phonological integration --------------------------------------------------------- Tightly integrated

Of course, things are a bit messier than this. Even some old agreement markers don't show obligatory doubling, especially object agreement markers which may be sensitive to pragmatics. But generally doubling becomes more frequent and more possible as the agreement markers erode and become more tightly integrated with their hosts. A form can move along the two dimensions above at different speeds, but the general direction along both is left to right.

[chris_notts]

On the topic of zero anaphora and control in "topic prominent" languages, I've gone looking for some examples to illustrate what I meant and found some in Anaphora by Huang. In one part of the book he contrasts "syntactic" languages (e.g. English) with "pragmatic" languages (e.g. Chinese, Japanese), in a way which I think correlates quite strongly with the subject vs topic prominent distinction.

First, a short quote:

Let me start with the pragmaticness of anaphora in a pragmatic language. If we compare a prototypical pragmatic language (such as Chinese, Japanese, and Korean) with a prototypical syntactic language (such as English, French and German), we will find that anaphora in the former behaves in a more pragmatic way than that in the latter (see e.g. Huang 1994 for a detailed analysis of Chinese). More specifically, a pragmatic language seems to have the following characteristics as far as anaphora goes.

Anaphora in a pragmatic language
(a) Massive occurrence of zero anaphora
(b) Existence of pragmatic zero anaphors or empty pragmatic categories
(c) Pragmatic obligatory control
(d) Long-distance reflexivization

The crucial point here for the previous discussion is that zeros are widely used, and not in a syntactically restricted way. Multiple zeros can occur in the same clause, with different antecedents, not restricted to be the marked topic of a previous or matrix clause.

Naomi-wa Ken-ni 0 0 aisiteiru-to itta
Naomi-TOP Ken-to 0 0 love-COMP say-PAST
"Naomi told Ken that (she) loved (him)"

John-un Mary-eykey cenhwa-lul hayssta kulikonun 0 0 0 salanghanta-ko malhayssta
John-TOP Mary-to telephone-ACC did and.then 0 0 0 love-COMP said
"John called Mary and then (he) said that (he) loves (her)"

Compare the following for an example where, in a relative clause, both zeros can be either the head or an external topic:

Laoshi hai zhao bu dao [yi ge [0 keyi jiao 0 de] xuesheng]
teacher still find not one CL 0 can teach 0 REL pupil
"The teacher still cannot find a pupil whom (he) can teach"

Xuesheng hai zhao bu dao [yi ge [0 keyi jiao 0 de] laoshi]
pupil still find not one CL 0 can teach 0 REL teacher
"The pupil still cannot find a teacher who can teach (him)"

In the following, a zero in a subordinate clause is controlled by the matrix subject (and topic) in the first example, and by the matrix recipient (and non-topic) in the other. The choice is pragmatically driven, not syntactic rule driven, except for the fact that some restrictions come from the choice of matrix verb "promise" (specifically, the controller can be the subject or recipient, but not easily by e.g. the matrix location):

Kanja-wa isya-ni asita 0 kusuri-o nomu koto o yakusokusi-ta
patient-TOP doctor-to tomorrow 0 medicine-OBJ take COMP OBJ promise-PAST
"The patient_1 promised the doctor_2 that (he_1) will take medicine tomorrow"

Isya-wa kanja-ni asita 0 taiinsuru koto o yakusokusi-ta
doctor-TOP patient-to tomorrow 0 leave hospital COMP OBJ promise-PAST
"The doctor_1 promised the patient_2 that (he_2) will leave the hospital tomorrow"

In these "topic prominent" languages, the fact that the subject is not the controller of zero anaphora does not mean that there is a replacement syntactic constraint that just replaces the word "subject" with "topic". Instead, the rules are a lot looser, and influenced not only by locally topic marked noun phrases, but referents that are discourse topical (even if they were last mentioned some time/distance ago), or just make sense in the context.

That's not to say that there aren't a small number of languages with something that looks like a grammaticalisation of a topic rule. I can think of a few languages which show things along those lines:

agreement with topics: Manambu
integration of topics into switch reference: Barai
integration of topicality into person paradigm: Algonquian languages (proximate)

But these are very different to the kind of topic prominence described by Li and Thompson for languages like Chinese and Japanese.

[bradrn]

Yes, maybe saying that clitics are never doubled is an exaggeration. In reality there's cline of obligatoriness and phonological integration:

Most clitic like ------------------------------------------------------------------------------- Most affix like
Not doubled ---------------------------------------------------------------------------------- Obligatory regardless of presence of coreferential noun phrase
Minimal phonological integration --------------------------------------------------------- Tightly integrated

Of course, things are a bit messier than this. Even some old agreement markers don't show obligatory doubling, especially object agreement markers which may be sensitive to pragmatics. But generally doubling becomes more frequent and more possible as the agreement markers erode and become more tightly integrated with their hosts. A form can move along the two dimensions above at different speeds, but the general direction along both is left to right.

This is pretty much what I was thinking as well.

On the topic of zero anaphora and control in "topic prominent" languages, I've gone looking for some examples to illustrate what I meant and found some in Anaphora by Huang. In one part of the book he contrasts "syntactic" languages (e.g. English) with "pragmatic" languages (e.g. Chinese, Japanese), in a way which I think correlates quite strongly with the subject vs topic prominent distinction. […]

This is pretty interesting! I guess I’ll take the zero anaphora approach in my conlang then.

[Pabappa]

Are there any nontrivial words ending in -tory in which the /o/ must be pronounced as a full vowel in all dialects of English? by "trivial" i mean I exclude words like Tory, story, and proper names.

Likewise, are there many words ending in -(a|i)tory in which the /o/ *cannot* be pronounced as a full vowel, even in AmE? I can think of some that end in just -tory, like refractory, but they all seem to have their /t/ coming from the root of the word rather than an affix like -at- or -it-.

[KathTheDragon]

I would always pronounce "auditory" with a non-reduced <o>

[Richard W]

Is story a trivial word?

Don't refrangible and refractory contain a common morpheme other than re-? The 't' you refer to is itself suffixal (or part of the -tor(-/y), depending on how you analyse it.

[chris_notts]

I would always pronounce "auditory" with a non-reduced <o>

Without the o, it would sound like it meant the art of auditing things.

[Richard W]

I would always pronounce "auditory" with a non-reduced <o>

Without the o, it would sound like it meant the art of auditing things.

English has a lot of Latinate homophones. These at least are different parts of speech, unlike the homophonous antonymic prefixes hypo- and hyper-.

[Pabappa]

Hmm. Thanks. The words with /ə/ in American English might just be explained by stress patterns, but auditory must have something else going on because there are dozens if not hundreds of words with that stress pattern that dont have an /o/.

Also, I always use a full /o/ in hypo-, so for me there is no confusion, and I think most people here, even if nonrhotic, still distinguish those two.

[zompist]

Likewise, are there many words ending in -(a|i)tory in which the /o/ *cannot* be pronounced as a full vowel, even in AmE? I can think of some that end in just -tory, like refractory, but they all seem to have their /t/ coming from the root of the word rather than an affix like -at- or -it-.

moratory, dormitory, expository, promontory, repository, celebratory, exploratory, observatory, conservatory.

(Some of these may have /o/ and /ə/ in free variation.)

[Travis B.]

Hmm. Thanks. The words with /ə/ in American English might just be explained by stress patterns, but auditory must have something else going on because there are dozens if not hundreds of words with that stress pattern that dont have an /o/.

Also, I always use a full /o/ in hypo-, so for me there is no confusion, and I think most people here, even if nonrhotic, still distinguish those two.

Remember that English has not just phonemic primary stress but phonemic secondary stress, which accounts for these cases.

[Nortaneous]

Likewise, are there many words ending in -(a|i)tory in which the /o/ *cannot* be pronounced as a full vowel, even in AmE? I can think of some that end in just -tory, like refractory, but they all seem to have their /t/ coming from the root of the word rather than an affix like -at- or -it-.

moratory, dormitory, expository, promontory, repository, celebratory, exploratory, observatory, conservatory.

(Some of these may have /o/ and /ə/ in free variation.)

huh? I've never heard any of those without a full vowel in AmE (except 'refractory', which I've never heard with one, and 'promontory', which I've never heard but which I think I'd pronounce with /ə/)