ANADEWs #2: The Finnish Object Case and its Complications

One sort of ANADEWy thing about Finnish is the case marking of the object, once you get down to the really fine-grained details. This post will first introduce the basic setup, and then go on with the worse bits.

The Finnish object case is determined in part by a semantic component, basically best illustrated by this schematic:

	telic	atelic
negative	part	part
affirmative	acc	part

Telicity is akin to perfectivity, but not quite the same thing. In school, I was taught the term "resultative" rather than telic. Basically, it is telic if these two semantic conditions hold:

• the action is talked of as achieving or having achieved the intended result, and
• the result of the action is the thing we're focusing on, not the progress of the action itself.
  

The accusative has a further complication, and this complication is syntactically conditioned. First of all, the nominative and accusative are identical in the plural, which means that basically, this decision algorithm is unnecessary there.

In the singular, the accusative is identical to the genitive. (However, some sources state this as 'there's an accusative I and an accusative II, the first being identical to the genitive, the other to the nominative.) If the verb is passive, or the subject is not in the nominative, the object will instead be in the accusative-nominative.

 

The main syntactic deciding factor is whether a VP can have a nominative subject. This might sound like a weird thing not to have, but there are a few types of constructions where nominative subjects are not possible in Finnish: several modal auxiliaries are quirky case verbs, and thus take the subject in the genitive. Imperatives cannot take nominative subjects. Passives cannot take nominative subjects. There are also some iffy bits about infinitives.

The decision flow for object case runs:

1. is the statement telic and affirmative?

yes? ↓       no? ↳ partitive

2. is it possible to have a nominative subject for that particular VP?

no? ↓   yes? ↳ accusative (genitive if singular, nom-acc if plural)

accusative (nominative if singular, nom-acc if plural)

The schematic above omits the whole question of infinitives, but by and large it is accurate. One could add an extra branch for plurals before 2, simplifying the reasoning for plurals but adding one extra step whenever it's a singular. I figure having a uniform decision algorithm is better.

 

Here are some examples:

Erkki osti auton
Erkki buy-past(3sg) car.gen
Since the result here - obtaining the car - is sort of the central point, it's telic, and thus some form of accusative. Since Erkki is an overt nominative subject,

Erkki ostaa auton
Erkki buy-3sg car.gen
Erkki (will) buy a car – the telicity implies that we're talking about him actually completing the transaction.

katselen televisiota
watch-1sg television-part
I am watching television

A further complication resides in the personal pronouns ̣- they all have an accusative form that is more similar to the plural nominative-accusative than to the genitive. Let's compare the noun morphology for nom, gen, acc1, acc2 and plur nom for the pronouns "minä", "sinä" (me, you) and the noun "talo" (house). The plural nom/acc of 1sg and 2sg does not really exist for personal pronouns in Finnish, as the plurals are formed from separate roots.

	nom	gen	"acc 1"	"acc 2"	plur nom/acc
1sg	minä	minun	minut	minut	n/a ("minut")
2sg	sinä	sinun	sinut	sinut	n/a ("sinut")
3pl	he	heidän	heidät	heidät	n/a ("heidät")
talo	talo	talon	talon	talo	talot

However, it turns out that somehow, the minut/sinut/... forms underlyingly are genitives or nominatives- they trigger genitive or nominal adjectival congruence depending on which object form is expected. The pronoun 'who', kuka, which is rich in suppletion (kuka, kene-, ke- being the principal roots) can have both kuka and kenet for "acc2" objects. The congruence with genitive adjectives also "feeds back" and forces the pronoun itself get genitive marking, and congruence with nominative adjectives operates likewise.

More on this stuff can be found in Paul Kiparsky's Structural Case in Finnish, a very thorough treatment of object and subject case in Finnish. 

What we as conlangers can take away from this is the potential for a case with limited distribution to behave weirdly outside of that context: an interesting thing would be to have more lexemes behave like kuka/kene-/ke- as per above. Alternatively, like the minut/sinut/... forms, the congruence could indicate an underlying case distinction.

Another option that I sort of envisioned while reading this was to have the minut/sinut/... gamut be underlyingly genitive throughout the system, and having adjectives marked for the "wrong case forms". By "wrong case forms", I mean as per the decision algorithm given above. So, when that decision algorithm indicates the nominative/acc2 case, but an adjective appears with a pronoun that conflates acc1 and acc2, the adjective appears in the genitive - and maybe even better, have the kuka/kene-/ke- -like thing happen and force the entire phrase into the genitive.

There's probably even more weird things one could come up with.

Pseudo-Numbers in Tarist

In Tarist, numbers and certain determiners form a closed class. Unlike in some languages, these differ significantly from adjectives.

The word order in the Tarist noun phrase is very much like in English, except adpositional attributes precede the head in adjectivified forms, subclauses often are replaced by participles, and adverbial attributes do not as such exist - adjectives are formed from adverbials instead. Unlike English, adpositional phrases can be either pre- or postpositional – this depends on the lexical properties of the adposition itself, as well as on syntactical factors - arguments are more likely to have prepositions than adjuncts are, for instance.

Given the above information, we can go on to this form:

[prep] [det] [numP] [adjP]* [noun] [relP]

Out of these, adjP is the only part that can follow a copula:

The X is adjP. It is an adjP X.

*The X are numP. They are numP X.
*The X is relP. They are X relP.

Normally, adjectives that are derived from adverbials and prepositional phrases lose their derivation when being extracted. Numeral phrases behave slightly differently - the subject will be in a quirky case, and the ~copula will be 'have' instead.

Numerals, unlike adjectives, affect the case and number marking of their head nouns: inanimate nouns after a numeral are in the singular. If the numeral is in the nominative, ergative or absolutive, the noun will be in the ablative (if neuter), or the ergative plural (if animate). (The obvious exception to this are numbers whose value is one or a non-integer or zero. Ones always are followed by the singular of the same case, rationals and zero by the ablative singular.)

Now for the lexical quirk - some words whose meaning we would consider more adjective-like are syntactically numerals in Tarist. This leads to some peculiarities. Examples:

bari - young
tars - old
knaedze - big, huge
xvurn - valuable, important
sogor - dead
xkuna - whole
muras - strong
sitvi - small

A peculiarity with these is that they can be part of big numerals. They can be inserted anywhere in a numeral construction, giving numbers like 'onehundredwholeteen', 'strongthousandfiftythree', etc. These adjectives too affect the case marking and the number marking of their heads (although for animates, they permit singular and plural marking - with inanimates, there are workarounds using other quantifiers in coordinated constructions).

Subjects and Case in Tarist

(Tarist is probably not going to be very developed; to the extent it fits into my language ideas more generally, it's closely related to Bryatesle, but has been influenced by some language family I've yet to come up with).

In Tarist, there are two kinds of subjects - proper subjects and improper subjects. The main visible distinction between the two is their position in the clause: proper subjects precede the verb, improper subjects follow it. A few other differences exist:

• all first and second person pronouns are proper subjects regardless of their position
• no verb congruence is triggered by improper subjects
• improper subjects do not take quirky case, but are invariably nominative
• proper subjects can bind reflexive anaphora, improper subjects cannot, and bind third person anaphora for reflexive utterances (and are thus slightly ambiguous)
• proper subjects have a mildly ergative case marking going on (except pronouns, which are strictly nom-acc)
• improper subjects are seldom topics, proper subjects are almost always topics
• improper subjects cannot be gapped over coordinated verbs, i.e. you can't do VERB SUBJ₁ (OBJ) and ____₁ VERB (OBJ), but you can do SUBJ₁ VERB (OBJ and ____₁ VERB (OBJ)
• passives only take improper subjects

Some verbs do require specifically one type or the other, but such verbs are few. Most verbs permit both, and which one uses depends, mostly, on the information structure of one's utterance. However, many nouns also lack case forms - a fair share lack the ergative, a fair share also lack the dative or the ablative. For verbs that force proper subjects to be marked with those cases, nouns that lack the relevant case will instead appear as an improper subject.

Detail #261: Definite Nouns as Less Semantically Precise

Let us consider a system whereby definite and indefinite nouns differ in a few ways. 

Let us consider the 'root', which might also be the definite form (or maybe the definite form takes some marking, but not particularly much). This needn't be very specific at all: context (i.e. the introduction of the referent) should be sufficient for the speaker to know what it is.

So we might have a lexeme meaning 'vessel'. This includes meanings such as 'ship', 'carriage', 'cart', 'boat', 'sleigh', 'chariot'. When first introducing the noun, some affix is required that specifies its type – and this is almost an open class in the language; there may be a 'generic' affix as well.

Another noun could be 'places in general', including suffixes turning it into 'house, burial ground, ritual place, village gathering place, pasture, ...'

Some nouns may be specific enough in their root form not to require any affixes - this at least applies to body parts, relations, certain types of locations, and certain animals and plants.

Outside of that, however, this system leads to certain problems: easily, many nouns whose definite forms are the same may coexist in a context. Therefore, the language would do well to have some kind of proximate-obviative system in place, or alternatively a definite-specific-indefinite system, where the specific type of noun does distinguish the same distinctions as the indefinite type?

Further, of course, this could combine with restrictions on the case system - the dative might not be permissible as indefinite, the genitive might not have an overt definite, but might be implicitly infinite anyways, the instrumental might lack the plural definite, the locative might not mark for definiteness at all, nor does it carry the distinctions.

Of course, the opposite way around – definite nouns as more semantically definite – could possibly also work.

Detail #260: A Noun Gender Marking System and Noun Possession

Let us consider a language with a two-gender system. The gender is intrinsic to the root, and is not marked morphologically in the basic form. Thus, you cannot predict the gender of a word from its spoken or written citation form.

There exist three markers, however, that are used when changing the gender, or under certain other situations. The markers are +masc, +fem, +poss/inv.

The gloss "poss/inv" may seem weird, but we'll get to that in a moment. A possessum will either be marked by '+poss/inv', if the possessor is the same gender as the possessum, or by the gender marker of the possessor's gender, if there's a difference. 

However, '+poss/inv' is also used to derive nouns that are the other gender. The derivation for 'agentive nouns', for most nouns, for instance, is masculine; adding +poss/inv, with no overt possessor, forms a feminine agent. Some verbs, however, do default to female agents, and for these, '+poss/inv' forms the male agent. Various animals also have default genders.

The basic outline here should also provide a framework to come up with even more convoluted possession- and gender-marking strategies.

Onwards with #258: Dummy Subjects for essentially intransitive verbs

A way of taking idea #258 further could be to permit intransitive subjects to become objects, if the subject of a previous, coordinable verb also is made the formal subject of the new verb phrase.

Anyways, let us assume there is no distinction between third person pronouns - he/her/it are all the same word, here represented as "E". The object form is 'em'.

John walks e's dog. E is happy.In this, John is happy.

John walks e's dog. Is-BIZARROVOICE happy em.
Here, John is not the one being happy, the dog is.

John walks e's dog. John/E_John is happy em_dog.
Here, again, the dog is happy. John serves only as a formal subject with no actual semantic relevance to the dog being happy. (Although, of course, the language could develop such connotations as well!)

John walks and is happy e's dog.Here, the verbs are coordinated; both are parsed as transitive, and therefore, it's the dog that is happy.

John walks his dog and is happy
Here, the verbs are coordianted. Since the object is close to the first verb, the only permissible parsing has the second verb intransitive, and therefore John being happy.

As for what it means for verbs to be coordinable, which as you may recall is a requirement I set up above, I left that somewhat unclear. Maybe the language only permits coordination of verbs of similar TAM? Maybe there are syntactical restrictions - some types of embedded verbs cannot be coordinated (e.g. a verb in a relative subclause cannot be coordinated with a verb in a matrix clause), etc. I leave this up to the interested conlanger to decide (although I might come up with ways of classifying verbs by 'coordinability' for conlanging purposes later on).

Detail #259: Splitting the Tense System in Two Parts

Tense systems come in a few different kinds, but two common, and clearly distinct types are the following ones:

	Past	Present	Future
past vs. non-past	past	non-past
future vs. non-future	non-future		future

An apology for my lack of an aesthetic sense with regards to colour would probably be justified about now.

One possibility is to have a conlang where both of these types appear, and even in the same contexts. Split the verbs into two "natural" classes, distinguished by how the action they depict reasonably interacts with tense.

Thus, habituals are likely to be of the past/non-past typ, as are non-dynamic states. Perfectives and more 'dynamic' actions, however, might be more likely to be of the non-future/future, etc. Some verbs might, for semantic reasons, be more likely to belong to one or the other - and there may be things like the nature of the subject or the nature of the object to influence the meaning.

The same morpheme could naturally mark non-future as well as non-past (with separate morphemes for past and future), or non-past and future vs.  non-future and past.

These are just some ramblings about possible things to do with the verbs in your language.