@cite{aissen-2003}: Differential Object Marking @cite{aissen-2003} #
Differential Object Marking: Iconicity vs. Economy. Natural Language & Linguistic Theory 21(3): 435–483.
Formalizes the core OT analysis: Harmonic Alignment of prominence scales with the relational scale (Subj > Obj) derives two constraint families:
- Iconicity (*Ø/X): penalizes zero-marked objects at prominence level X. Fixed ranking: *Ø most prominent >>... >> *Ø least prominent.
- Economy (*!/X): penalizes marked objects at prominence level X. Fixed ranking: *!/least prominent >>... >> *!/most prominent.
Rankings are fixed within each family but free between families. The factorial typology over all consistent interleavings predicts exactly the attested DOM patterns.
Key Results #
| Scale Size | Consistent Rankings | Language Types | Impossible Patterns |
|---|---|---|---|
| 2 elements | 6 | 3 | Mark low without high |
| 3 elements | 20 | 4 | Any non-monotone pattern |
For the 3-element animacy scale {Hu > An > In}, 4 of 8 logically possible patterns are generated — exactly the monotone ones (Table 17, p. 476).
Connection to Existing Infrastructure #
The predicted DOM profiles are matched against the DOMProfile language data
in §0 below, verifying that every attested pattern corresponds
to a possible OT grammar.
A DOM (Differential Object Marking) profile: a DifferentialMarkingProfile
specialized to role P + channel flagging.
Each cell (a, d) records whether an object with animacy level a
and definiteness level d obligatorily receives an overt DOM marker
(e.g., Spanish a, Turkish -(y)I, Hindi -ko).
DOM is the P-flagging instance of @cite{just-2024}'s general differential
marking framework. Monotonicity (isMonotone), isAnimacyOnly, and
isDefinitenessOnly are all inherited from DifferentialMarkingProfile.
Instances For
Spanish: a-marking for human direct objects regardless of definiteness.
One-dimensional (animacy-based), cutoff between human and animate.
Equations
Instances For
Russian: animate accusative (genitive form used as accusative for animate nouns). One-dimensional (animacy-based), cutoff between animate and inanimate.
Equations
Instances For
Turkish: -(y)I marking for definite direct objects regardless of
animacy. One-dimensional (definiteness-based), cutoff between definite
and indefinite specific.
Equations
Instances For
Hebrew: ʔet marking for definite direct objects regardless of
animacy. Same one-dimensional definiteness cutoff as Turkish.
Equations
Instances For
Persian: -rā marking for definite direct objects. One-dimensional
(definiteness-based) for obligatory marking; optional extension to
specific indefinite animates.
Equations
Instances For
Catalan: a-marking restricted to personal pronouns. The most
restrictive DOM pattern attested.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Hindi-Urdu: -ko marking conditioned by BOTH animacy and definiteness.
Two-dimensional DOM with a staircase cutoff.
Equations
- One or more equations did not get rendered due to their size.
Instances For
No DOM: trivially monotone. Included in allDOMProfiles to match OT
Type 4 (no marking) via nodom_matches_type4 in §6.
Equations
- One or more equations did not get rendered due to their size.
Instances For
All DOM profiles in the sample (7 attested + 1 no-marking baseline).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Aissen's DOM monotonicity universal: all attested DOM patterns in the sample form upper sets in the bidimensional animacy × definiteness grid. No language marks a less-prominent object while leaving a more-prominent one unmarked.
Hindi DOM depends on both animacy and definiteness — it cannot be reduced to a single scale.
In all sample languages, human objects are never less marked than animate objects at the same definiteness level.
In all sample languages, animate objects are never less marked than inanimate objects at the same definiteness level.
In all sample languages, pronouns are never less marked than proper names at the same animacy level.
In all sample languages, definite NPs are never less marked than indefinite specific NPs at the same animacy level.
If any cell is marked, the most prominent cell (human, pronoun) is also marked.
The least prominent cell (inanimate, non-specific) is unmarked in all DOM languages in the sample.
The bidimensional grid has 3 × 5 = 15 cells per language.
Total marked cells across all sample languages.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Marked cells: Spanish (5) + Russian (10) + Turkish (9) + Hebrew (9) + Persian (9) + Catalan (3) + Hindi (7) + NoDOM (0) = 52.
All interleavings of two lists, preserving internal order of each.
Given two constraint families with fixed internal rankings, this generates all total orders consistent with both. The number of interleavings of lists of lengths m and n is C(m+n, m).
Equations
- One or more equations did not get rendered due to their size.
- Aissen2003.interleavings [] x✝ = [x✝]
- Aissen2003.interleavings x✝ [] = [x✝]
Instances For
DOM candidate for a 2-element prominence scale {High > Low}.
true = overtly case-marked; false = zero-marked.
- high : Bool
- low : Bool
Instances For
Equations
- Aissen2003.instDecidableEqScale2Cand.decEq { high := a, low := a_1 } { high := b, low := b_1 } = if h : a = b then h ▸ if h : a_1 = b_1 then h ▸ isTrue ⋯ else isFalse ⋯ else isFalse ⋯
Instances For
Equations
- Aissen2003.instReprScale2Cand = { reprPrec := Aissen2003.instReprScale2Cand.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Instances For
*Ø/High: penalize unmarked High objects.
Equations
- Aissen2003.starZeroHigh = Core.Constraint.OT.mkMark "*Ø/High" fun (c : Aissen2003.Scale2Cand) => c.high = false
Instances For
*Ø/Low: penalize unmarked Low objects.
Equations
- Aissen2003.starZeroLow = Core.Constraint.OT.mkMark "*Ø/Low" fun (c : Aissen2003.Scale2Cand) => c.low = false
Instances For
*!/Low: penalize marked Low objects (economy).
Equations
- Aissen2003.starBangLow = Phonology.Constraints.mkDep "*!/Low" fun (c : Aissen2003.Scale2Cand) => c.low = true
Instances For
*!/High: penalize marked High objects (economy).
Equations
- Aissen2003.starBangHigh = Phonology.Constraints.mkDep "*!/High" fun (c : Aissen2003.Scale2Cand) => c.high = true
Instances For
Iconicity family (fixed: *Ø/High >> *Ø/Low).
Instances For
Economy family (fixed: *!/Low >> *!/High).
Equations
Instances For
All consistent rankings: interleavings of the two families.
Instances For
There are exactly 6 consistent rankings (C(4,2) = 6).
Compute optima for each consistent ranking.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The 2-element scale yields exactly 3 language types, not 4 (Table 14, p. 473).
The impossible pattern — mark Low without High — is never optimal.
DOM candidate for the 3-element animacy scale {Hu > An > In}.
true = overtly case-marked; false = zero-marked.
- hu : Bool
- an : Bool
- inan : Bool
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
Equations
- Aissen2003.instReprAnimCand = { reprPrec := Aissen2003.instReprAnimCand.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Iconicity: *Ø/Hu >> *Ø/An >> *Ø/In.
Equations
- Aissen2003.starZeroHu = Core.Constraint.OT.mkMark "*Ø/Hu" fun (c : Aissen2003.AnimCand) => c.hu = false
Instances For
Equations
- Aissen2003.starZeroAn = Core.Constraint.OT.mkMark "*Ø/An" fun (c : Aissen2003.AnimCand) => c.an = false
Instances For
Equations
- Aissen2003.starZeroIn = Core.Constraint.OT.mkMark "*Ø/In" fun (c : Aissen2003.AnimCand) => c.inan = false
Instances For
Economy: *!/In >> *!/An >> *!/Hu.
Equations
- Aissen2003.starBangIn = Phonology.Constraints.mkDep "*!/In" fun (c : Aissen2003.AnimCand) => c.inan = true
Instances For
Equations
- Aissen2003.starBangAn = Phonology.Constraints.mkDep "*!/An" fun (c : Aissen2003.AnimCand) => c.an = true
Instances For
Equations
- Aissen2003.starBangHu = Phonology.Constraints.mkDep "*!/Hu" fun (c : Aissen2003.AnimCand) => c.hu = true
Instances For
Iconicity family (fixed: *Ø/Hu >> *Ø/An >> *Ø/In).
Equations
Instances For
Economy family (fixed: *!/In >> *!/An >> *!/Hu).
Equations
Instances For
All consistent rankings for the 3-element scale.
Equations
Instances For
There are exactly 20 consistent rankings (C(6,3) = 20).
Compute optima for each consistent ranking.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Distinct language types.
Equations
- Aissen2003.animTypes = Aissen2003.animOptima.eraseDups
Instances For
The 3-element animacy scale yields exactly 4 language types, not 8 (Table 17, p. 476).
Every generated type is monotone: if An is marked then Hu is too; if In is marked then An is too (Aissen's central prediction).
Type 1: mark all (Hu + An + In). Extreme iconicity.
Type 2: mark Hu + An only. Russian pattern (animate accusative).
Type 3: mark Hu only. Spanish pattern (personal a).
Type 4: mark none. No DOM (economy dominates).
Mark In without An: never generated.
Mark An without Hu: never generated.
Mark In without Hu: never generated.
Mark In only (without An or Hu): never generated.
Convert an AnimCand to a one-dimensional animacy DOMProfile.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Every OT-generated animacy type produces a monotone DOMProfile.
Spanish DOM (human only) matches OT Type 3 (Hu only).
Russian DOM (animate+) matches OT Type 2 (Hu + An).
No-DOM profile matches OT Type 4 (none marked).
The 2-element definiteness scale {Pro > D} from §4 of the paper, where "D" covers definite NPs (including proper names). This gives the same 3-type factorial typology as any 2-element scale.
Convert a Scale2Cand to a definiteness-based DOMProfile. High = personalPronoun, Low = properName + definite (i.e., ≥ definite).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Catalan DOM (pronouns only) matches 2-element Type 2 (High only).
Turkish DOM (definite+) matches 2-element Type 1 (both marked).
Hebrew DOM (definite+) also matches 2-element Type 1.
Of 8 logically possible 3-element patterns, OT generates exactly 4.
Of 4 logically possible 2-element patterns, OT generates exactly 3.
The number of consistent rankings grows as C(2n, n). For n=2: C(4,2) = 6. For n=3: C(6,3) = 20.
An NP enriched with referential prominence properties.
Structural case assignment (dependent case) is blind to these properties — it cares only about c-command and lexical case. DOM then consults prominence to decide overt realization.
- label : String
- lexicalCase : Option Core.Case
- animacy : Features.Prominence.AnimacyLevel
- definiteness : Features.Prominence.DefinitenessLevel
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
Equations
- Aissen2003.instReprProminentNP = { reprPrec := Aissen2003.instReprProminentNP.repr }
Strip prominence, yielding the NP that the case algorithm sees.
Equations
- pnp.toNP = { label := pnp.label, lexicalCase := pnp.lexicalCase }
Instances For
A transitive clause: subject c-commands object.
- subject : ProminentNP
- object : ProminentNP
Instances For
Equations
- Aissen2003.instDecidableEqTransClause.decEq { subject := a, object := a_1 } { subject := b, object := b_1 } = if h : a = b then h ▸ if h : a_1 = b_1 then h ▸ isTrue ⋯ else isFalse ⋯ else isFalse ⋯
Instances For
Equations
- Aissen2003.instReprTransClause = { reprPrec := Aissen2003.instReprTransClause.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Run the dependent case algorithm on a transitive clause.
Equations
- Aissen2003.derivation lang tc = Syntax.Case.assignCases lang [tc.subject.toNP, tc.object.toNP]
Instances For
Abstract case assigned to the object.
Equations
- Aissen2003.objectCase lang tc = Syntax.Case.getCaseOf tc.object.label (Aissen2003.derivation lang tc)
Instances For
Whether the object receives overt case morphology.
Two conditions:
- The dependent case algorithm assigns ACC (syntax).
- The DOM profile marks this prominence cell (morphology).
Equations
- Aissen2003.objectOvert lang dom tc = (Aissen2003.objectCase lang tc == some UD.Case.acc && dom.marks tc.object.animacy tc.object.definiteness)
Instances For
A standard transitive clause with a fixed subject (human pronoun) and a variable-prominence object. Both lack lexical case.
Equations
- One or more equations did not get rendered due to their size.
Instances For
In accusative transitives, the object receives abstract ACC regardless of its animacy or definiteness. Dependent case is prominence-blind.
The subject always gets NOM (unmarked case).
The overt marking profile produced by running the full pipeline (dependent case + DOM filter).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Every OT-predicted animacy type, run through the full pipeline, produces a monotone overt marking profile.
The pipeline preserves monotonicity for all attested DOM languages.
All 8 attested DOM profiles, run through the accusative case pipeline, produce overt marking that is faithful to the DOM input AND monotone.