@cite{bruening-2025} — Selectional Violations in Coordination: A Response

Bruening, Benjamin. 2025. Selectional violations in coordination: A response to Patejuk and Przepiórkowski 2023. Linguistic Inquiry 56(3). 439–483.

Three AMT Experiments

Experiment 1a (n=65): Adverbs vs Adjectives in Prenominal Position

Tests whether non-ly adverbs can appear in prenominal coordination with adjectives. The adverb condition is significantly degraded relative to the adjective condition (t=7.28, p<.001) but more acceptable than ungrammatical baselines, consistent with Adv↔Adj being a permitted but marked selection violation.

Experiment 1b (n=65, same participants): CPs in NP Positions

Tests the coordination rescue effect: CPs in DP-selecting positions are more acceptable in coordination with DPs than alone. Coordination condition significantly better than bare CP (t=5.575, p<.001).

Experiment 2 (n=77): One-Replacement Test

Tests whether one-replacement accepts adverbs in coordination. Adjective one-replacement is significantly more acceptable than adverb one-replacement (t=6.895, p<.001).

C-Selection Irreducibility (§4.1)

become/grow/get/end up/turn out have identical semantics (change-of-state to property state) but different c-selectional profiles, proving c-selection is not reducible to s-selection.

structure Bruening2025.ExpCondition : Type

Condition-level descriptive statistics for AMT acceptability experiments. Z-scores stored as Int × 1000 (milli-z) for exact arithmetic.

• label : String
• meanZ : ℤ
• sdZ : ℤ

def Bruening2025.instDecidableEqExpCondition.decEq (x✝ x✝¹ : ExpCondition) : Decidable (x✝ = x✝¹)

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance Bruening2025.instDecidableEqExpCondition : DecidableEq ExpCondition

Equations

• Bruening2025.instDecidableEqExpCondition = Bruening2025.instDecidableEqExpCondition.decEq

@[implicit_reducible]

instance Bruening2025.instReprExpCondition : Repr ExpCondition

Equations

• Bruening2025.instReprExpCondition = { reprPrec := Bruening2025.instReprExpCondition.repr }

def Bruening2025.instReprExpCondition.repr : ExpCondition → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance Bruening2025.instBEqExpCondition : BEq ExpCondition

Equations

• Bruening2025.instBEqExpCondition = { beq := Bruening2025.instBEqExpCondition.beq }

def Bruening2025.instBEqExpCondition.beq : ExpCondition → ExpCondition → Bool

Equations

• Bruening2025.instBEqExpCondition.beq { label := a, meanZ := a_1, sdZ := a_2 } { label := b, meanZ := b_1, sdZ := b_2 } = (a == b && (a_1 == b_1 && a_2 == b_2))
• Bruening2025.instBEqExpCondition.beq x✝¹ x✝ = false

def Bruening2025.exp1a_nParticipants : ℕ

Experiment 1a: 65 AMT masters participants. Tests prenominal coordination [Adj and Adv] N. Data from Table 2 of @cite{bruening-2025}.

Equations

• Bruening2025.exp1a_nParticipants = 65

def Bruening2025.exp1a_control : ExpCondition

Equations

• Bruening2025.exp1a_control = { label := "Control (grammatical)", meanZ := 630, sdZ := 710 }

def Bruening2025.exp1a_ungram : ExpCondition

Equations

• Bruening2025.exp1a_ungram = { label := "Ungrammatical", meanZ := -970, sdZ := 940 }

def Bruening2025.exp1a_adj : ExpCondition

Equations

• Bruening2025.exp1a_adj = { label := "Adjective coordination", meanZ := 870, sdZ := 400 }

def Bruening2025.exp1a_adv : ExpCondition

Equations

• Bruening2025.exp1a_adv = { label := "Adverb coordination", meanZ := -310, sdZ := 830 }

theorem Bruening2025.exp1a_adj_better_than_adv : exp1a_adj.meanZ > exp1a_adv.meanZ

Adjective coordination is more acceptable than adverb coordination.

theorem Bruening2025.exp1a_adv_intermediate : exp1a_adv.meanZ > exp1a_ungram.meanZ ∧ exp1a_adv.meanZ < exp1a_control.meanZ

Adverb coordination is rated between grammatical and ungrammatical: not fully acceptable, but not as bad as outright ungrammaticality. This is consistent with adverb-as-adjective being a permitted but degraded selection violation (@cite{bruening-alkhalaf-2020} §3.2).

def Bruening2025.exp1a_tValue : ℤ

Experiment 1a paired t-test (Adj vs Adv): t=7.28, df=8.35, p<.001.

Equations

• Bruening2025.exp1a_tValue = 7280

def Bruening2025.exp1a_df : ℤ

Equations

• Bruening2025.exp1a_df = 8350

def Bruening2025.exp1b_nParticipants : ℕ

Experiment 1b: same 65 participants as Exp 1a. Tests CP in DP-selecting position: coordination vs simple. Data from Table 6 of @cite{bruening-2025}.

Equations

• Bruening2025.exp1b_nParticipants = 65

def Bruening2025.exp1b_control : ExpCondition

Equations

• Bruening2025.exp1b_control = { label := "Control (grammatical)", meanZ := 630, sdZ := 710 }

def Bruening2025.exp1b_ungram : ExpCondition

Equations

• Bruening2025.exp1b_ungram = { label := "Ungrammatical", meanZ := -970, sdZ := 940 }

def Bruening2025.exp1b_coord : ExpCondition

Equations

• Bruening2025.exp1b_coord = { label := "DP-CP coordination", meanZ := 280, sdZ := 670 }

def Bruening2025.exp1b_simple : ExpCondition

Equations

• Bruening2025.exp1b_simple = { label := "Simple CP", meanZ := -500, sdZ := 740 }

theorem Bruening2025.exp1b_coord_rescues : exp1b_coord.meanZ > exp1b_simple.meanZ

DP-CP coordination is more acceptable than bare CP: the coordination rescue effect. A CP that cannot appear alone in a DP-selecting position becomes acceptable when coordinated with a DP.

theorem Bruening2025.exp1b_simple_degraded : exp1b_simple.meanZ < exp1b_control.meanZ

Bare CP is rated below grammatical control: it IS unacceptable without coordination rescue.

theorem Bruening2025.exp1b_coord_above_ungram : exp1b_coord.meanZ > exp1b_ungram.meanZ

Coordination rescue raises acceptability above the ungrammatical baseline.

def Bruening2025.exp1b_tValue : ℤ

Experiment 1b paired t-test (Coord vs Simple): t=5.575, df=7.984, p<.001.

Equations

• Bruening2025.exp1b_tValue = 5575

def Bruening2025.exp1b_df : ℤ

Equations

• Bruening2025.exp1b_df = 7984

def Bruening2025.exp2_nParticipants : ℕ

Experiment 2: 77 AMT masters participants. Tests one-replacement in [Adj and Adv] coordination. Data from Table 4 of @cite{bruening-2025}.

Equations

• Bruening2025.exp2_nParticipants = 77

def Bruening2025.exp2_filler_gram : ExpCondition

Equations

• Bruening2025.exp2_filler_gram = { label := "Filler (grammatical)", meanZ := 750, sdZ := 670 }

def Bruening2025.exp2_filler_ungram : ExpCondition

Equations

• Bruening2025.exp2_filler_ungram = { label := "Filler (ungrammatical)", meanZ := -780, sdZ := 870 }

def Bruening2025.exp2_adj : ExpCondition

Equations

• Bruening2025.exp2_adj = { label := "Adjective one-replacement", meanZ := 360, sdZ := 630 }

def Bruening2025.exp2_adv : ExpCondition

Equations

• Bruening2025.exp2_adv = { label := "Adverb one-replacement", meanZ := -490, sdZ := 720 }

theorem Bruening2025.exp2_adj_better_than_adv : exp2_adj.meanZ > exp2_adv.meanZ

Adjective one-replacement is more acceptable than adverb one-replacement.

theorem Bruening2025.exp2_adv_intermediate : exp2_adv.meanZ > exp2_filler_ungram.meanZ ∧ exp2_adv.meanZ < exp2_filler_gram.meanZ

Adverb one-replacement is less acceptable than grammatical fillers but more acceptable than ungrammatical fillers.

def Bruening2025.exp2_tValue : ℤ

Experiment 2 paired t-test (Adj vs Adv): t=6.895, df=7.67, p<.001.

Equations

• Bruening2025.exp2_tValue = 6895

def Bruening2025.exp2_df : ℤ

Equations

• Bruening2025.exp2_df = 7670

theorem Bruening2025.shared_controls : exp1a_control.meanZ = exp1b_control.meanZ ∧ exp1a_ungram.meanZ = exp1b_ungram.meanZ

Experiments 1a and 1b share the same participants and filler items. The identical control/ungrammatical baselines confirm this.

Five change-of-state verbs have identical semantics (transition to property state) but different c-selectional profiles (§4.1):

• become: AP ✓, NP ✓ ("became rich", "became a doctor")
• grow (copular sense): AP ✓ only ("grew tired"; distinct from agentive grow "grow potatoes" which takes NP)
• get: AP ✓, PP ✓ ("got tired", "got into trouble")
• end up: AP ✓, PP ✓ ("ended up tired", "ended up in trouble")
• turn out: CP ✓ ("It turned out that...")

Since s-selection is identical (all select for a property state), the variation must be c-selectional. This proves c-selection is irreducible to s-selection, contra @cite{przepiorkowski-2024} who argue that category mismatches in coordination reflect semantic rather than syntactic selection.

structure Bruening2025.CSelProfile : Type

C-selectional profile of a change-of-state predicate.

• verb : String
• takesAP : Bool
• takesNP : Bool
• takesPP : Bool
• takesCP : Bool

@[implicit_reducible]

instance Bruening2025.instDecidableEqCSelProfile : DecidableEq CSelProfile

Equations

• Bruening2025.instDecidableEqCSelProfile = Bruening2025.instDecidableEqCSelProfile.decEq

def Bruening2025.instDecidableEqCSelProfile.decEq (x✝ x✝¹ : CSelProfile) : Decidable (x✝ = x✝¹)

Equations

• One or more equations did not get rendered due to their size.

def Bruening2025.instReprCSelProfile.repr : CSelProfile → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance Bruening2025.instReprCSelProfile : Repr CSelProfile

Equations

• Bruening2025.instReprCSelProfile = { reprPrec := Bruening2025.instReprCSelProfile.repr }

@[implicit_reducible]

instance Bruening2025.instBEqCSelProfile : BEq CSelProfile

Equations

• Bruening2025.instBEqCSelProfile = { beq := Bruening2025.instBEqCSelProfile.beq }

def Bruening2025.instBEqCSelProfile.beq : CSelProfile → CSelProfile → Bool

Equations

• One or more equations did not get rendered due to their size.
• Bruening2025.instBEqCSelProfile.beq x✝¹ x✝ = false

def Bruening2025.become : CSelProfile

Equations

• Bruening2025.become = { verb := "become", takesAP := true, takesNP := true, takesPP := false, takesCP := false }

def Bruening2025.grow_ : CSelProfile

Equations

• Bruening2025.grow_ = { verb := "grow", takesAP := true, takesNP := false, takesPP := false, takesCP := false }

def Bruening2025.get_ : CSelProfile

Equations

• Bruening2025.get_ = { verb := "get", takesAP := true, takesNP := false, takesPP := true, takesCP := false }

def Bruening2025.endUp : CSelProfile

Equations

• Bruening2025.endUp = { verb := "end up", takesAP := true, takesNP := false, takesPP := true, takesCP := false }

def Bruening2025.turnOut : CSelProfile

Equations

• Bruening2025.turnOut = { verb := "turn out", takesAP := false, takesNP := false, takesPP := false, takesCP := true }

def Bruening2025.changeOfStateVerbs : List CSelProfile

Equations

• Bruening2025.changeOfStateVerbs = [Bruening2025.become, Bruening2025.grow_, Bruening2025.get_, Bruening2025.endUp, Bruening2025.turnOut]

theorem Bruening2025.become_turnOut_complementary : become.takesAP = true ∧ become.takesCP = false ∧ turnOut.takesAP = false ∧ turnOut.takesCP = true

turn out and become have complementary c-selection profiles despite identical semantics: become takes AP/NP but not CP, turn out takes CP but not AP/NP.

theorem Bruening2025.cselection_varies : ¬(changeOfStateVerbs.all fun (x : CSelProfile) => x == become) = true

Not all five verbs share the same c-selection profile.

theorem Bruening2025.get_endUp_same_profile : get_.takesAP = endUp.takesAP ∧ get_.takesNP = endUp.takesNP ∧ get_.takesPP = endUp.takesPP ∧ get_.takesCP = endUp.takesCP

get and end up pattern together (both take AP and PP).

theorem Bruening2025.experiments_confirm_two_violations : BrueningAlKhalaf2020.permittedViolations.length = 2 ∧ exp1b_coord.meanZ > exp1b_simple.meanZ ∧ exp1a_adv.meanZ > exp1a_ungram.meanZ

All three experiments confirm B&AK's two-violation restriction: CP↔NP (Exp 1b) and Adv↔Adj (Exp 1a, 2) are the only permitted selection violations in coordination.

theorem Bruening2025.exp1b_tests_cpAsNp : BrueningAlKhalaf2020.permittedViolations.contains BrueningAlKhalaf2020.SelectionViolationType.cpAsNp = true ∧ BrueningAlKhalaf2020.SelectionViolationType.cpAsNp.cats = (Core.Tree.Cat.CP, Core.Tree.Cat.NP)

The coordination rescue effect in Exp 1b is exactly the CP↔NP violation type: a CP that cannot appear alone in a DP-selecting position becomes acceptable when coordinated with a DP. Grounded in coordExtension: NP ∈ coordExtension CP.

theorem Bruening2025.exp1a_tests_advAsAdj : BrueningAlKhalaf2020.permittedViolations.contains BrueningAlKhalaf2020.SelectionViolationType.advAsAdj = true ∧ BrueningAlKhalaf2020.SelectionViolationType.advAsAdj.cats = (Core.Tree.Cat.AdvP, Core.Tree.Cat.AdjP)

The prenominal adverb effect in Exp 1a is the Adv↔Adj violation type: a non-ly adverb in an adjective-selecting position. Grounded in coordExtension: AdjP ∈ coordExtension AdvP.