Sometimes systems in languages can be analyzed in terms of feature bundles. This can work for cases, tenses, moods, lexical subsystems (say, family terminology, or such), etc. This article will look into the case systems of several of my languages, attempting to find some economical yet powerful description of the case systems.
Ŋʒädär has a fairly simple case system:
The absolute marks subjects and objects. The dative marks recipients. The genitive-comitative marks possessors or accompanying participants. The three locatives - locative, lative and ablative - have similarities, but there is an odd one out among them. The complement case has certain similarities to the absolute case.
We find beyond these that there is an unabsolutive case for certain nouns. Counting the regular cases we have seven, and the unabsolutive would give us eight. log28 is 3, so the most optimal case would only have three binary variables. Let us first look at the seven 'common' cases before taking a look at the unabsolutive.
It seems three basic qualities distinguish the three cases: involvement, direction and centrality. Involvement is whether the noun is a participant in any way whatsoever, or just a frame or scaffold for the action. Direction is whether there is a spatial progress involving the noun as some form of source or sink, and 'centrality' largely corresponds to likelihood of being topicalized or focalized but also the likelihood of being an argument.
The question marks indicate that the relevant spots seem to go both ways. The ablative thus can acquire the same meaning as the lative in some contexts, but can also acquire a distinct meaning. We can expand this by having both the complement and the ablative appear as two versions of themselves - giving a total of nine, but this is ok since ablative2 is the same as lative as far as its features go.
here, the cases are ordered assuming participant > central > directional
By now we have exhausted the number of states three binary variables can occupy, so the unabsolutive wouldn't fit into this. We could attempt to rearrange this so that we get rid of the question marks and express both the complement and ablative in terms that do not require them to occupy two different states - however, this particular setup will prove useful to understand the shenanigans of the case systems. We shall rearrange the system a bit for a truly full three-variable system without any cases occupying two slots, and using a different set of features that better catch the "morphological reality". The middle column has different values for the upper and lower half.
This model also has its drawbacks; 'active core case' signifies cases that (can) participate in an action, but obviously the absolutive can be the object as well, and quirky case verbs can take datives that do things. Framing is a question of locating a VP or subject either spatially or conceptually. Associate reference is whether a noun in such a case necessarily refers to a noun's referent itself or possibly to things associated therewith - i.e. the locative may be marked upon a noun the vicinity of which is referred to (whereas the lative more usually goes on the name of a place, or a noun on the inside or top of which something moves).
Asking what features an NP satisfies for these two schemes gives a pretty good idea of what case an NP in Ŋʒädär takes, but even then the two models give some mistakes. Similar models for Ćwarmin would be huge, but Bryatesle, Sargaĺk and Dairwueh may get their own treatment among these lines. On the other hand, the interaction of number, case and definiteness in Ćwarmin could make for interesting models that demonstrate how weirdly intervowen those three really are in Ćwarmin.