@cite{amato-2025} — Agreement case studies via Nested Agree #
@cite{sigurdsson-holmberg-2008}
@cite{amato-2025} (NLLT) advances Nested Agree — formalized at
Theories/Syntax/Minimalism/NestedAgree.lean — as a unifying account
across at least six syntactic phenomena. This file collects the
case studies whose primary phenomenon is agreement:
- §4.1.2 — Icelandic DAT-NOM intervention (
DatNom): quirky-dative subjects with nominative objects exhibit optional agreement with the lower nominative across the structurally higher dative. Derived from feature ordering on T (@cite{sigurdsson-holmberg-2008}). - §4.2.1 — Lak perfective person agreement (
Lak): in the Nakh-Daghestanian language Lak, perfective aspect forces person agreement on the lowest argument (the absolutive), even across the unmarked ergative subject. Derived from a Nested Agree configuration on Asp with cyclic Agree threading the φ-features from the absolutive through v.
The Italian auxiliary selection chapter (§3) lives at
Phenomena/AuxiliaryVerbs/Studies/Amato2025.lean (different primary
phenomenon). Cross-domain cases beyond agreement — Bulgarian
multiple wh-fronting (§4.2.3), ditransitive case alignments (§4.1.1)
— are deferred to future companion files in their respective
phenomenon directories.
Cross-domain validation thesis #
Every case study here uses the same NestedAgreeConfig shape with
zero API changes — different SyntacticObjects, different choices
of goalHead, different validGoal predicates, but one Theory layer.
This validates NestedAgree.lean as a genuine cross-domain primitive
rather than Italian-shaped scaffolding.
Related infrastructure #
Phenomena/Case/Studies/Marantz1991.lean contains an independent
treatment of dative intervention via DativeInterventionContext and
dativeIntervenes. Bridge theorems connecting Marantz's threshold
analysis to Amato's feature-ordering analysis are future work.
Part A: Icelandic DAT-NOM intervention (Amato §4.1.2) #
Icelandic quirky dative subjects with nominative objects exhibit a well-known optionality. @cite{amato-2025} §4.1.2 derives this from two co-existing feature orderings on T (@cite{sigurdsson-holmberg-2008}):
- Ordering A:
[*case:nom*] ≻ [*φ:_*]— case-Agree fires first and selects the nominative DP (skipping the case-marked dative subject); φ-Agree by Nested Agree must target the same goal, yielding agreement with the nominative. - Ordering B:
[*φ:_*] ≻ [*case:nom*]— φ-Agree fires first and hits the dative subject (closest), but the dative is φ-defective for valuation purposes; default 3sg agreement surfaces.
The two orderings give two NestedAgreeConfigs with the same root
SyntacticObject but different goalHead choices and validGoal
predicates. Optionality follows from which is well-formed.
Out of scope: biclausal Icelandic (Amato 29b, requires modeling clause embedding); the Sigurðsson-Holmberg A/B/C dialect taxonomy (this section proves optionality holds in Icelandic A; B/C variation is a follow-up).
Configuration A — agreement with the nominative #
Ordering A: [*case:nom*] ≻ [*φ:_*]. Case-Agree skips the
case-marked dative and reaches the nominative; φ-Agree by Nested
Agree must target the same nominative goal. All heads are
phi-active under this ordering — the dative is irrelevant since
it's not the chosen goal. Tree: T [DPdat [V DPnom]], goal = DPnom.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ordering A is well-formed: T c-commands DPnom in icelandicTree,
and DPnom is phi-active.
Apparent dative intervention is not actual. The dative subject is in T's c-command (probe 0's domain) but is not in DPnom's daughters (DPnom doesn't c-command DPdat — DPdat is structurally above DPnom). DPdat is excluded from probe 1's truncated domain by Nested Agree.
Configuration B — default agreement #
Ordering B: [*φ:_*] ≻ [*case:nom*]. φ-Agree fires first and hits
the dative subject (closest), but the dative's φ-features are
quirky/defective for finite T's valuation purposes.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ordering B is not well-formed: the chosen goal is φ-defective. The formal expression of "default agreement surfaces because π-Agree on T fails to value."
Optionality #
Monoclausal optionality (Amato 29a, Sigurðsson-Holmberg 2008 Icelandic A): the same surface tree admits both feature orderings, so both agreement outcomes are predicted. The optionality is a property of T's feature ordering, not of the underlying structure.
Both orderings operate on the same root tree — the optionality is in the probe stack's feature ordering, not in the structural representation.
Bridge to @cite{marantz-1991}'s threshold analysis #
@cite{marantz-1991} models dative intervention through a *case
accessibility threshold*: a dative DP intervenes (blocks the probe)
when its lexical case is below the probe's accessibility threshold
(so the probe can't Agree with it) but it still blocks access to
the intended target by minimality. The Marantz analysis produces
a Bool `dativeIntervenes`; @cite{amato-2025} §4.1.2 produces a
Prop `¬ IsNestedAgreeConfig orderingB`. The bridge below shows
they make the same prediction on the same configuration.
A DativeInterventionContext corresponding to ordering B's
structural setup: dative present, standard unmarked threshold
(lexical case is not accessible), nominative target.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Bridge: under ordering B's parameter setting,
@cite{marantz-1991}'s dativeIntervenes predicts intervention iff
@cite{amato-2025}'s IsNestedAgreeConfig rejects the configuration.
Both theories agree: dative intervention → agreement failure
(default 3sg in Amato's surface, agreement-failure in Marantz's
threshold model).
Part B: Lak perfective person agreement (Amato §4.2.1) #
Lak is a Nakh-Daghestanian language with ergative-absolutive case marking. In the present tense, person agreement is controlled by the external argument if it's not ergative-marked (1st/2nd person under the differential ergativity pattern of @cite{radkevich-2017}); else by the absolutive object. In the perfective aspect, however, person agreement is always controlled by the lowest argument (the absolutive), even when the external argument is unmarked (@cite{amato-2025} §4.2.1, examples 31a–d).
Amato derives this from Nested Agree on the perfective Asp head:
Asp bears [*Infl:perf*] ≻ [*π:_*]. Infl-Agree fires first and
agrees with v (creating the channel); π-Agree by Nested Agree must
target the same v. v has prior π-Agreed with the absolutive object
via cyclic Agree (Legate 2005), so v carries the absolutive's
φ-features. Result: agreement on Asp surfaces as the absolutive's
φ-features, across the unmarked ergative.
Structurally identical to the Italian transitive aux selection
configuration (Italian: Perf [DPsubj [v DPobj]] → goal v;
Lak: Asp [Erg [v Abs]] → goal v). The same NestedAgreeConfig
shape applies — what changes is the linguistic interpretation of
the surface output.
Out of scope: non-perfective Lak (no Asp head, no Nested Agree configuration, standard agreement controlled by case-marking threshold per @cite{marantz-1991}-style accessibility); the contrast between (31a/b) and (31c/d) demonstrating the perfective's distinctive pattern.
The Lak perfective configuration. Tree: Asp [Erg [v Abs]],
goal = v. v is reachable (c-commanded by Asp) and phi-active
(carries Abs's φ via cyclic Agree).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Apparent ergative intervention is not actual. The ergative subject is in Asp's c-command (probe 0's domain) but is not in v's daughters (v doesn't c-command Erg — Erg is structurally above v in Spec,vP). Erg is therefore excluded from probe 1's truncated domain.
This is @cite{amato-2025} §4.2.1's structural resolution of the Lak agreement puzzle: the ergative looks like a closer goal but Nested Agree restricts probe 1 to v's subtree, where Erg has no presence. Surface result: agreement with Abs (transitively through v's prior cyclic Agree).