Detail #124: Topics and Case Marking

It seems to me that topics would be a natural place for the case marking rules in a language to have any number of exceptions - and I even imagine these exceptions could be both towards less marking or more marking.

One idea could be to conflate all case marking on topics with the nominative, except that the accusative is retained for objects of a certain nature, and some very specific verb phrases retain specific markings - say the language has a Finnish/Russian-like predicative possession construction, maybe that particular case is retained as well in that particular construction.

Another idea could be to have, say, differential marking on subjects or objects - but only when they're topics. (So, non-topic objects are exclusively accusative, but as topics they can also be some other case - nominative or dative or whathaveyou, subjects as non-topics are exclusively nominative, but for topical subjects, the genitive gives definite subjects).

The justification why case marking could be less detailed follows from the example of Chinese, where topics often have no adpositions. The intuition we could have for this is that the topic is a sort of obvious participant in some way, and we are more likely to be able to expect its role in the sentence than other constituents' roles. However, on the other hand, topics could also imaginably attract more marking, as very central constituents in clauses.

Small Musing on Prepositions Governing Case

Most conlangers who are probably aware of how in several languages - Latin, German, Russian as prime examples - prepositions alter meaning depending on which case they are used with. In Indo-European languages, dative often goes with location, and accusative often with direction. Alternatively, in say Russian, there's the prepositional and instrumental cases that both are paired with the accusative to the same effect.

However, we're also aware that word order can distinguish subjects from objects and other similar distinctions. In addition, we know other things can distinguish subject from object - lexical knowledge about which noun out of two is more likely to be the subject in general, which noun is more likely to be the subject or object of which verb, etc.

Compare how in English, 'at' can be quite locative at times, and quite directional at other times. In part this seems to correlate* with whether it's a syntactic complement of the verb or not, more locative in nature the less complementy it is. (Here, I am thinking of complement in the way that the object is a syntactic complement.) 

So, maybe we could have the difference between a preposition being locative or directional as a result of word order in the sentence - S Prep N O V = S at N O V, S O Prep N V = S O towards N V.

And to make it better, turn this into a statistical likelihood, i.e. O Prep N is more likely to mean 'towards' than to mean 'at', but that there still is a significant probability for the meaning to be 'at'.

* Have not checked this and won't check it. It's purely a gut feeling, and I am not going to base what essentially is a speculation about how a language might work on whether this gut feeling is correct or not.

Dairwueh: The Third Person Verb

The Dairwueh verb, for most tense-aspect combinations have two third person forms, labelled quite clearly 3rd person I and 3rd person II. Since the II-form does not distinguish number, the following indexing is practical: 3_sg, 3_pl, 3_II. The use of the 3_II form requires some elucidation.

Generally, 3_II is formed with the least actual morphology out of the third person forms. If the verb, inflected for tense, ends in a bimoraic syllable, that is that is also the 3_II form. However, if the last syllable is monomoraic, it gains a second mora in some way.

It is used in the following circumstances:

• with non-explicit subjects
• with impersonal constructions
• non-topical subjects
• relative subclauses (where the subject is relativized), and other subclauses where the subject is not present in that particular subclause)
• on verbs that are not the primary predicate of the clause
• sometimes with indefinite subjects

Non-explicit subjects include 'pseudopassives', i.e. just not having a subject.

hāg - (people) do.PRESENT-3_II (that)
(from hag, 'to do')

erŋe sawnī - thing-PL.ACC buy-PAST.3_II, things were bought, the things got sold

It does not include pronoun dropping though.

In relative subclauses, it is possible for the subject to be present outside of the clause but not within it. In such cases, this too blocks the regular third person marking in the subclause. Non-primary predicates tend to mark meanings along the lines of certain English adverbs. These might have an impersonal subject or the same subject as the main verb, but are marked the same regardless whether impersonal or not.

Indefinite subjects are more likely to have this form the less topical they are.

Detail #123: One idea for Alignment and Causatives

So, we know how coordination should  under syntactic ergativity:

girl_erg slapped boy_abs and fell

In syntactic ergativity, it is the boy that falls. In most ergative languages that also permit coordination, that is not the case, indicating they have a nominative-accusative syntax underlying the case marking. However, let's do some fun stuff with this and causatives.

We have two primary types of construction:

S O V 

S V 

Now, let's assume we're having a nom-acc language for now.
We get causative transformations thus:

C S O V

C O V

C S V

The C O V instance is obviously somewhat causative-passive, i.e. C causes someone to verb O. We would probably have marking along these lines:

C_nom S_{some case 1?} O_{some case 2?} V

C_nom  O_{some case 2?} V

C_nom  S_{some case 1?} V

It is possible S is marked as some kind of non-object,  and 1 and 2 might reasonably be the same. Let's now consider what happens when we have coordinate this with some other verb. Remember, non-causatives here coordinate like in English: S_nom O_acc V1 and V2 means Snom did V1 and V2, regardless of transitivity of V2.

 C_nom S_{some case 1?} felled and fell

 This could easily be a way of marking telicity of causatives! However, let's look at the other situation

 C_nom S_{case 1} O_{case 2} V_1,trans and V_2,intr

 John_nom made Eric_{case 1} paint the house_{case 2} and turned red : the house turned red (due to S1's painting it)

 C_nom O_{case 2} V_1,trans and V_2,intr

 C_nom made the house be painted and turned red: the house turned red (due to C's making people paint it)

We do get this kind of situation though: 

 C_nom S_{case 1} O_{1,case 2} V_1,trans and O_2,acc V_2,trans 

in pseudo-English with SVO order instead: 

 John_nom made Eric_{case 1} paint the house_{case 2} and spilled the paint_acc

This could reasonably go both ways:  John_nom or Eric_{case 1} could both be the subjects of V_2,trans from a purely a posteriori viewpoint. We could leave such a thing ambiguous, or we could restrict it to either of the two, or we could have some peculiar semantically empty 'pseudo-voice' that is used to resolve this, i.e. a 'decausative voice' that is used exclusively to coordinate your regular verbs with a causer, or maybe the other way around. We also have the situation where we may want to coordinate an intransitive verb with either the C or S argument. We could also have the case marking on S1 depend on whether we want it to be the subject of an embedded verb - maybe shifting it rightwards or marking it as nominative if we are going to coordinate with it?

Detail #122: Places to put Alignmenty Things

[Some edits have been done due to feedback on the clarity. Great thanks to an American gentleman for helping me out]

Alignment-related stuff is an interesting way of showing off just how non-average non-European your language is. However, it is all too easy to describe a language-encompassing alignment and leave it at that - or maybe specify some rules for when the language aligns with nominative-accusative and when it goes ergative-absolutive.

There are more things to do with alignment than just that though! An obvious bit is in participles and derivative morphology - English famously has 'escapee' following an ergative pattern, but that is not the only possibility. A Siberian language whose name escapes me at the moment has its negative participles follow an ergative pattern, whereas the rest of the language is strictly accusative. 

So, let's get onto some weird places to put some exceptional alignments. Let's assume primarily that we're making an ergative language where accusative alignments keep popping up everywhere. Other combinations of alignments could be made just as well. 

1. Causatives

One could easily have the causer's case marking follow an ergative pattern. In these glosses, C marks the causer, and indices mark the case they mark:

We start with two quite regular sentences:

S_erg O_abs V_trans

 S_abs V_intr

 Adding the causer, C, we obtain, as now the S is notionally an object if no embedded object is present, and using obl as an embedded ergative:

C_erg S_abs V_caus,intr 

C_erg S_obl O_abs V_caus,trans

But much more variety would be achieved with nominative marking in an otherwise ergative language, and this version is what we'll develop further:

C_erg S_abs V_caus,intr 

C_erg S_abs O_obl V_caus,trans

Of course, there's also the possibility of omitting the embedded subject, and just saying that Subject caused Object to be Verbed, in which case we can have anti-ergative patterns, here using the somewhat unclear notion previous encountered in the post on 'intransitive objects' :

C_erg O_abs V_caus,trans 

C_erg S_abs V_caus,trans

or a nominative pattern:

_Cerg O_abs V_caus,trans (causee omitted)
_Cerg S_obl V_caus,trans (object omitted)

Ergative patterns basically do exist for this kind of setup, but they would kind of be odd given that we'd be inserting an ergative pattern into a nominative pattern in an ergative pattern:

_Cerg O_obl V_caus,trans 

C_erg S_obl V_caus,trans

2. Participles

We can have passive participles of transitive verbs marked the same as active participles of intransitive verbs in a nominative language, thus giving us an ergative subsystem. However, we could also have nominative-style participles in an ergative language, thus giving us basically the same kinds of participles that English have. However, we could also have this system break down in some constructions, so that it reverts to the more typical system under certain circumstances - such as, say, with some auxiliary or maybe when the participle is a dangling participle. 

Of course, lexical exceptions also enable funny stuff.

3.  Other non-finite verbal subject and object marking?

4. Some secondary thing, like having possession of subjects marked in one way, and possessions of objects in some other way, in an otherwise very ergative language, or alternatively possession of intransitive subjects and objects marked the same but otherwise a very nominative language.

5. Some other participant marking? (I.e. having the presence of marking for the listener affect cases of the verb arguments? Having marking for the evidentiality-source affect the case marking, in ways that are reminiscent of alignment? This is an idea I came up with a minute before posting this, so will need some further development, and it might turn out that it's ill-conceived from the beginning.)

6. To really make this baroque, one could of course have different noun classes further divide up the case space differently, so that humans follow clearly nominative tendencies both with regular verbs and with causatives (and other similar constructions), whereas other noun classes are nominative, but antiergative in the causatives, some are antiergative, some are ergative with regular verbs but nominative with causatives, etc. Such an alignments galore language would be weird, though, and probably not very realistic. There's probably some hierarchy for what behavior in the regular transitive verb can coexist with what alignment between the arguments of the causative verb, and what kinds of ditransitive alignments are likely to coexist with what kinds of transitive alignments - and even further probably some implicational universals between the causative and ditransitive alignments.

Detail #121: Adverbs as Complements

Imagine a language where adverbs generally are seen either as a complement of the subject (i.e., complements as in 'I am angry', 'I turned weak', etc, but generalized so any verb can have it - 'I walked quick' (where quick describes 'me', rather than 'walked') rather than 'I walked quickly', 'I fought convinced') or the object (as in the type of complement you get in 'I painted my house red'). Now, let's do this by having the adverbs agree with their subject or object in case and or gender.

Now, obviously, some adverbs don't necessarily have any clear semantic connection to either subject or object - 'he definitely knows our plan', 'regrettably, I cannot tell you this secret'. Thus, the language gets a lot of lexicalized information with each adverb-like thing as to whether they agree with the subject or the object. In the case of intransitive sentences, though, they all agree with the subject. (Alternatively, they agree with some default gender/case?) (Actually, this could be a place to insert some split-ergativity!)

Of course, even locative expressions that are not core arguments of the verb could go for some congruence-thing like this.

Ćwarmin, Bryatesle and Dairwueh: Some Basic Word Order Typology

Ćwarmin is an SOV language, although some flexibility is present. It does follow some tendencies for such languages:

• postpositions
• (Det) (Num) (Adj) Nom, 
• Gen N - although like in most of my conlangs, Gen can be extracted in several ways
• wh- in situ
• auxiliaries follow infinitives
• a tendency towards time-manner-place order in adverbs

Ćwarmin has "historically" developed from an intermediate SVO stage that developed out of a previous SOV stage, so there are a handful of SVO-like features clinging on, although it never thoroughly acquired full SVO compliance. Relative clauses can precede or follow the noun, although there's a more restricted set of ways in which they can be formed before the noun.

Bryatesle too is basically SOV, with the same basic word order tendencies:

• postpositions
• (Det) (Num) (Adj) Nom
• Gen N, although again, the gen can be extracted in several ways
• wh- in situ
• auxiliaries follow infinitives
• a tendency towards time-manner-place order in adverbs

Bryatesle has no traces from any non-SOV systems. Both Ćwarmin and Bryatesle permit OSV and the verb can be fronted as well. SVO and OVS are very uncommon in both, although an extra subject pronoun can appear in Bryatesle after the verb for emphasis. 

Dairwueh, on the other hand, is SVO. It has the following properties:

• prepositions
• Nom (Num) (Det) (Adj) 
• N Gen
• prevalent wh-movement

Both Dairwueh and Ćwarmin have the complication that many verbs that to a speaker of English might seem to serve auxiliary-like roles are not really auxiliaries at all, but rather what we might term 'pragmatic verbs', verbs which serve as building blocks for information structure rather than as a ways of signalling actions or states of participants, and these tend to be coordinated with other verbs rather than being in an auxiliary verb-main verb relationship. In Bryatesle, this is rather dealt with by means of a huge set of conjunctions and discourse particles.