• case-driven suppletion
• case syncretism
• case cartography
• dependent case

In this thesis, I attempt to lay out a comprehensive survey of *ABA patterns in case
morphology—that is, patterns whereby a certain realizational rule can never apply to
both case x and case z without also applying to case y. I show, more specifically, that such *ABA patterns can be found in systematic case syncretism, in case-conditioned wholesale suppletion, and in case-conditioned stem-formative allomorphy. I also argue that, crucially, the universal *ABA patterns evidenced by all these phenomena consistently make reference to the very same case classes, as encapsulated in (i).
(i) a. If a systematic syncretism covers both an unmarked core case (nominative or absolutive) and an inherent case (dative, instrumental, locative, etc.), it will also cover a marked core case (accusative or ergative).
b. If the same stem formative appears in an unmarked core case (nominative or absolutive) and in an inherent case (dative, instrumental, locative, etc.), it will also appear in a marked core case (accusative or ergative).
c. If the same root appears in an unmarked core case (nominative or absolutive) and in an inherent case (dative, instrumental, locative, etc.), it will also appear in a marked core case (accusative or ergative).
The generalizations in (i) trace back to recent work by Baerman et al. (2005) and Caha (2009) on syncretism, by McFadden (2014) on stem allomorphy, and by Smith et al. (2016) on suppletion. Most of hese works, however, albeit largely consistent with each other, invoke some or other distinctive qualification or supplementation to the general pattern. Here I argue, in contrast, that such ‘deviations’ are not adequately empirically supported, and that the match between the relevant patterns across the three phenomena at issue is indeed perfect or nearly so. This argumentation rests, among other things, on a systematic defence of (i-a) against a sample of over 100 diverse case-inflecting languages, as well as on a sustained critique of Caha’s (2009) proposals concerning several other putative *ABA universals in syncretisms with or among inherent cases.
Next, drawing again on Caha (2009) and Smith et al. (2016), I interpret the patterns
summarized in (i) as evidence for a containment hierarchy like (ii), according to which
every inherent case contains a marked core case, and every marked core case in turn
contains an unmarked one.
(ii) [[[ nom/abs ] acc/erg ] inherent ]
This hierarchy, while totally unexpected under theories that treat the ergative as just another inherent case (e.g. Woolford 1997; Legate 2008), can by contrast be naturally implemented under those theories that have the ergative assigned in much the same way as the accusative. Prominent among these approaches is Marantz’s (1991) dependent-case theory, whereby accusative and ergative are both treated as dependent cases—i.e. assigned to nominals which stand in an asymmetric c-command relation to some other non-oblique nominal nearby. In this light, I propose to reformulate (ii) as (iii).
(iii) [[[unmarked] dependent] inherent]
Finally, I observe that the lightest cases in the containment hierarchy (nominative
and absolutive) are also the same cases that are given default (or, more accurately, ‘elsewhere’) status in the Marantzian theory of case assignment. This naturally prompts the conjecture that there might be a structural economy principle behind this pattern—a
principle whereby additional case layers such as those of dependent and inherent cases
may only be deployed in case of necessity, when the lighter, unmarked cases are somehow
‘not enough’. The question arises, then, why these latter are sometimes ‘not enough’.
The answer I will tentatively explore will be to suppose that clauses must try their best to abide by the two requirements in (iv).
(iv) a. Selectional requirements must be satisfied.
b. Two noun phrases that stand in an asymmetric c-command relation within the same domain must bear sufficiently distinct cases (cf. Richards 2010).
On this account, the additional case structure of ergatives and accusatives may be deployed only when needed to tell apart two otherwise–overly-similar noun phrases, while
the case structure associated to obliques may be deployed only if specifically selected by some idiosyncratic lexical item.
I conclude by discussing a range of problems besetting such an account, including
some of the difficulties surrounding the proper treatment of tripartite and active~inactive alignments.