@cite{de-hoop-malchukov-2008}: Case-Marking Strategies @cite{de-hoop-malchukov-2008} #
@cite{aissen-2003} @cite{blutner-2000}
Case-Marking Strategies. Linguistic Inquiry 39(4): 565–587.
Derives the basic case-marking typology from two functional constraints and an economy constraint, evaluated in @cite{blutner-2000}'s Bidirectional OT:
- Identify (I): Case identifies the role — overt marking should match argument strength (markedness principle). Penalizes form/strength mismatch.
- Distinguish (D): Case distinguishes A from P — unmarked atypical arguments are confusable with the other role. Penalizes zero-marked atypical arguments.
- Economy (*!) / PaIP: Avoid overt case marking (general) or specifically protect the primary actant from marking (alignment-sensitive).
Bidirectional OT #
Standard OT selects the best form for a given meaning (speaker). BiOT adds
the hearer's perspective via superoptimality (@cite{blutner-2000} eq. 14,
"weak optimality"): a form–meaning pair ⟨f, m⟩ is superoptimal iff no other
superoptimal pair with the same form or same meaning is strictly more harmonic.
This is a greatest-fixed-point computation
(Core.ConstraintEvaluation.superoptimal).
Under weak BiOT, the 2×2 form-meaning game (2 forms × 2 meanings) always produces Horn's division of pragmatic labour regardless of ranking — the fixed-point computation re-admits marked-form / marked-meaning pairs whose blockers were themselves eliminated. This means Economy/PaIP-dominant rankings give the same results as functional-constraint-dominant rankings.
Key Results #
1. DSM divergence (subjects, §2) #
Identify and Distinguish make opposite predictions for subjects:
| Ranking | Superoptimal pairs | Marking pattern |
|---|---|---|
| I >> *! | (ERG, strong), (Ø, weak) | Mark strong subjects (Manipuri) |
| D >> *! | (Ø, strong), (ERG, weak) | Mark weak subjects (Fore) |
| *! >> I/D | → Ø | Neutral (no DSM) |
2. DOM convergence (objects, §2) #
Identify and Distinguish make the same prediction for objects:
| Ranking | Superoptimal pairs | Marking pattern |
|---|---|---|
| I >> *! | (ACC, strong), (Ø, weak) | Mark strong objects |
| D >> *! | (ACC, strong), (Ø, weak) | Mark strong objects |
3. Symmetrical ⇒ Identify (§3) #
When a language has two overt case forms (e.g., Finnish ACC/PART), Distinguish is vacuously satisfied by all overt pairs — it cannot drive the choice between them. Symmetrical marking is necessarily Identify-driven (p. 574–578).
4. PaIP and alignment correlation (§4) #
The Primary Actant Immunity Principle (PaIP) restricts which argument can be differentially marked. When PaIP dominates:
- In nom-acc: subject = primary actant → PaIP >> I/D blocks DSM → only DOM
- In ergative: object = primary actant → PaIP >> I/D blocks DOM → only DSM
This derives: DOM ↔ nom-acc, DSM ↔ ergative (p. 580).
Case forms for a single argument: zero or overt marking. Used for asymmetrical differential marking (overt alternates with Ø).
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Argument strength: prominence relative to role prototypicality.
"Strong" means prominent (animate, definite, pronominal). The crucial insight is that strength interacts differently with each role:
- Strong A = prototypical agent (inherently distinguishable from P)
- Strong P = atypical patient (confusable with A)
This reversal is why Distinguish targets weak subjects but strong objects, while Identify uniformly targets strong arguments.
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Identify (I): Case should identify the argument's role. Overt marking matches strong arguments (marking = marked role); zero matches weak arguments (no marking = unmarked role). Penalizes form/strength mismatch: (overt, weak) and (zero, strong).
Same for both argument positions — Identify is role-independent. Tableau (18) and (29).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Distinguish for subjects (D_A): Unmarked weak subjects are confusable with objects. Strong subjects are inherently distinguishable. Only (Ø, weak) violates.
Weak A is atypical (agent-like properties are low) → confusable with P. Strong A is typical → clearly agent, no marking needed. Tableau (21), p. 573.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Distinguish for objects (D_P): Unmarked strong objects are confusable with subjects. Weak objects are inherently distinguishable. Only (Ø, strong) violates.
The reversal from distinguishSubj is the key to DOM convergence:
strong patients are atypical (agent-like), so Distinguish targets
them — the same pattern as Identify.
Tableau (32)/(35).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Economy (*!): Penalizes any overt case marker. Same as @cite{aissen-2003}'s economy family.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Identify and Distinguish make opposite predictions for subjects. This divergence explains the two attested DSM strategies.
I >> *! for subjects: strong subjects get ERG (Manipuri pattern). Tableau (18), p. 572: Identify drives overt marking of the prototypical (strong) agent, because ERG "identifies" the strong agent role.
D >> *! for subjects: weak subjects get ERG (Fore pattern). Tableau (21), p. 573: Distinguish drives overt marking of the atypical (weak) agent, because it's confusable with the patient.
*! >> I for subjects: under weak BiOT, ranking-independent — same result as I >> *!. The fixed-point re-admits ⟨overt, strong⟩ once its blocker ⟨zero, strong⟩ is eliminated.
*! >> D for subjects: under weak BiOT, ranking-independent — same result as D >> *!.
Identify and Distinguish make the same prediction for objects: mark strong (prominent) patients. This convergence is the paper's central result (p. 576) — it explains the universality and robustness of DOM compared to the variability of DSM.
The convergence arises because "strong" objects are atypical patients:
Identify wants to mark them (overt = strong), and Distinguish also
wants to mark them (strong P is confusable with A).
I >> *! for objects: strong objects get ACC. Tableau (31): ACC identifies the prominent patient.
D >> *! for objects: strong objects get ACC — same as Identify! Tableau (32)/(35): ACC distinguishes the prominent (confusable) patient from the agent.
DOM convergence (p. 576): Identify and Distinguish produce identical superoptimal sets for objects. This is the strongest form: the literal output lists are equal.
DSM divergence: Identify and Distinguish produce different
superoptimal sets for subjects. Contrast with dom_convergence.
Extract the case form assigned to a given strength level from the superoptimal set. If no pair matches, defaults to zero.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A case-marking pattern for one argument position: what form each strength level receives.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Derive the marking pattern from a ranking's superoptimal set.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Subject typology (Table 1, p. 573). Under weak BiOT, ranking is irrelevant for the 2×2 game: Economy-dominant rankings produce the same patterns as constraint-dominant rankings.
Object marking always targets strong regardless of ranking (under weak BiOT). Both I and D agree on the object pattern, and the pattern is ranking-independent.
Languages like Finnish (ACC vs. PART for objects) and Lezgian (ERG vs. OBL for subjects) use two overt case forms that alternate by argument strength. This is symmetrical differential marking.
The key theorem: Distinguish cannot derive symmetrical marking. All
overt forms satisfy Distinguish equally, so Distinguish has no basis
for assigning different overt forms to different strengths. The result
is that both overt forms map to the same strength (strong for objects,
leaving weak uncovered). Only Identify can pair each overt form with
a distinct strength.
"asymmetrical differential case marking can sometimes be explained by
DISTINGUISHABILITY, symmetrical differential case marking is necessarily
due to IDENTIFY" (p. 576).
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Identify for 3-form system: form1 matches strong, form2 matches weak. Tableau (29), p. 576: ACC identifies strong P, PART identifies weak P.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Distinguish for objects in 3-form system: only (Ø, strong) violates. Both overt forms satisfy Distinguish identically.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Economy for 3-form system.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Distinguish cannot discriminate overt forms (structural theorem). Both overt forms receive identical violations from Distinguish for any meaning. This is the formal basis for the claim that symmetrical marking requires Identify.
I >> *! in 3-form system: form1 for strong, form2 for weak. Tableau (29), p. 576: Finnish ACC/PART alternation. Each overt form is paired with a distinct strength.
D >> *! in 3-form system: both overt forms map to strong, zero maps to weak. Distinguish cannot differentiate form1 from form2 — both satisfy D identically — so both are superoptimal for strong. No overt form is paired with weak.
Symmetrical marking requires Identify: under Identify, each overt form is paired with a unique strength (form1↔strong, form2↔weak). Under Distinguish, no overt form is paired with weak — the symmetrical alternation cannot arise.
The Primary Actant Immunity Principle (PaIP) replaces Economy with an alignment-sensitive constraint. The primary actant is the argument encoded like the intransitive subject: - In nominative-accusative: the A/S subject - In ergative-absolutive: the P/S object
PaIP penalizes overt marking of the primary actant. When PaIP outranks
the functional constraints (I, D), it blocks differential marking of
the primary actant. This derives:
- DOM is found in nom-acc (marking objects doesn't violate PaIP)
- DSM is found in ergative (marking subjects doesn't violate PaIP)
"DOM is characteristic of nominative-accusative languages, while DSM
is found primarily in ergative languages" (p. 580).
When I/D conflicts with PaIP (e.g., Identify wants to mark the primary
actant), voice alternation resolves the conflict: passive in nom-acc,
antipassive in ergative (p. 582).
PaIP: Penalizes overt marking of the primary actant. Same violation profile as Economy — the difference is that PaIP only applies to the primary actant's position.
Equations
- One or more equations did not get rendered due to their size.
Instances For
PaIP is ranking-independent under weak BiOT (p. 580). PaIP has the same violation profile as Economy, so PaIP >> I/D produces the same result as I/D >> Economy. The paper discusses PaIP-I/D conflict as resolved through voice alternation (passive in nom-acc, antipassive in ergative), not through OT tableaux.
Nom-acc: DOM proceeds (p. 580). Objects are NOT the primary actant, so PaIP doesn't apply. I >> *! for objects: mark strong patients (DOM).
Ergative: DSM proceeds (p. 580). Subjects are NOT the primary actant, so PaIP doesn't apply. D >> *! for subjects: mark weak agents (DSM).
Alignment correlation (p. 580). Under weak BiOT, both argument positions show differential marking regardless of ranking. The asymmetry between nom-acc (DOM but not DSM) and ergative (DSM but not DOM) is not derivable from ranking alone in the 2×2 game — the paper resolves this via voice alternation (passive/antipassive) when PaIP and I/D conflict.
Economy (*!) is the same constraint across both analyses. The DOM convergence result (§4) explains why Aissen's harmonic alignment works for DOM: both the Identify and Distinguish motivations predict the same monotone pattern.
Aissen's iconicity family (*Ø/X) corresponds to the Distinguish
perspective: marking targets atypical arguments first. For P, this
aligns with Identify (both target strong P). For A, Distinguish
diverges from Identify (targeting weak A vs. strong A).
Economy has the same violation profile as Aissen's economy family: 1 violation per overt marker, 0 for zero markers.
DOM convergence is consistent with Aissen's monotonicity: overt marking targets the top of the prominence hierarchy first. Under both Identify and Distinguish, strong (prominent) objects are marked before weak (non-prominent) ones.