Minimal Pairs — Theory-Neutral Judgment Data Vocabulary

Shared vocabulary for introspective grammaticality contrasts — the minimal-pair tradition in which a datum is a pair of minimally-different sentences with a per-side acceptability judgment (typically the analyst's). This is the methodological tradition that the factorial-design machinery of [Spr07] / [SWP12] (Linglib/Studies/SprouseEtAl2012.lean) was introduced to discipline.

Two parallel families:

• Word-based (MinimalPair, PhenomenonData): sentences are List Word, requiring feature specifications. Used by analyses that operate on parsed/featural representations (HPSG, DG, Minimalism).
• String-based (SentencePair, StringPhenomenonData): sentences are raw strings, parseable by any theory. Used by data files that should remain free of theoretical commitments.

Judgments use the five-level Features.Judgment scale (Linglib/Features/Acceptability.lean).

Word-based

structure Features.MinimalPairs.MinimalPair : Type

A minimal pair: two candidate forms with per-side acceptability judgments. The judgment fields default to the traditional binary (lhs = .acceptable, rhs = .ungrammatical) so most data sites only need to provide lhs and rhs; studies of graded phenomena (e.g. CRDC marginality, [OL23]) override the judgments with the appropriate Judgment constructor.

• lhs : List Word
• rhs : List Word
• lhsJudgment : Judgment
• rhsJudgment : Judgment
• clauseType : ClauseForm
• description : String
• citation : String

def Features.MinimalPairs.instReprMinimalPair.repr : MinimalPair → ℕ → Std.Format

Equations

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

@[implicit_reducible]

instance Features.MinimalPairs.instReprMinimalPair : Repr MinimalPair

Equations

• Features.MinimalPairs.instReprMinimalPair = { reprPrec := Features.MinimalPairs.instReprMinimalPair.repr }

structure Features.MinimalPairs.PhenomenonData : Type

Collection of minimal pairs for a phenomenon.

• name : String
• pairs : List MinimalPair
• generalization : String

@[implicit_reducible]

instance Features.MinimalPairs.instReprPhenomenonData : Repr PhenomenonData

Equations

• Features.MinimalPairs.instReprPhenomenonData = { reprPrec := Features.MinimalPairs.instReprPhenomenonData.repr }

def Features.MinimalPairs.instReprPhenomenonData.repr : PhenomenonData → ℕ → Std.Format

Equations

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

def Features.MinimalPairs.predictionAgrees (predicted : Bool) : Judgment → Bool

Whether a Bool-valued grammaticality prediction agrees with a Judgment. Categorical predicates can only adjudicate the endpoints of the scale; for graded judgments (.marginal, .questionable, .unacceptable) the predicate is treated as vacuously consistent, since a Bool theory is not equipped to make those distinctions.

Equations

• Features.MinimalPairs.predictionAgrees predicted Features.Judgment.acceptable = predicted
• Features.MinimalPairs.predictionAgrees predicted Features.Judgment.ungrammatical = !predicted
• Features.MinimalPairs.predictionAgrees predicted x✝ = true

def Features.MinimalPairs.capturesMinimalPairBy (pred : List Word → Bool) (pair : MinimalPair) : Bool

Check if a grammaticality predicate captures a minimal pair.

Captures iff the predicate's Bool result agrees with each side's Judgment per predictionAgrees. With default judgments this is the traditional pred lhs && !pred rhs.

Equations

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

def Features.MinimalPairs.capturesPhenomenonData (pred : List Word → Bool) (phenom : PhenomenonData) : Bool

Check if a grammaticality predicate captures all minimal pairs in a phenomenon dataset.

Equations

• Features.MinimalPairs.capturesPhenomenonData pred phenom = phenom.pairs.all (Features.MinimalPairs.capturesMinimalPairBy pred)

String-based (theory-neutral)

structure Features.MinimalPairs.SentencePair : Type

String-based minimal pair for theory-neutral phenomena.

Unlike MinimalPair which uses List Word (requiring feature specifications), this type uses raw strings that can be parsed by any theory. This keeps empirical data free from theoretical commitments.

• grammatical : String

  The grammatical sentence

• ungrammatical : String

  The ungrammatical sentence

• clauseType : ClauseForm

  Clause form (declarative, question, etc.)

• description : String

  Description of what the pair tests

• citation : String

  Citation for the data; empty string for uncited examples.

@[implicit_reducible]

instance Features.MinimalPairs.instReprSentencePair : Repr SentencePair

Equations

• Features.MinimalPairs.instReprSentencePair = { reprPrec := Features.MinimalPairs.instReprSentencePair.repr }

def Features.MinimalPairs.instReprSentencePair.repr : SentencePair → ℕ → Std.Format

Equations

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

def Features.MinimalPairs.instBEqSentencePair.beq : SentencePair → SentencePair → Bool

Equations

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

@[implicit_reducible]

instance Features.MinimalPairs.instBEqSentencePair : BEq SentencePair

Equations

• Features.MinimalPairs.instBEqSentencePair = { beq := Features.MinimalPairs.instBEqSentencePair.beq }

structure Features.MinimalPairs.StringPhenomenonData : Type

String-based phenomenon data for theory-neutral representation.

This is the string-based analogue of PhenomenonData. Data files should use this type so that empirical data is decoupled from any particular grammatical theory's representation.

• name : String

  Name of the phenomenon

• pairs : List SentencePair

  List of minimal pairs

• generalization : String

  Generalization captured by this data

@[implicit_reducible]

instance Features.MinimalPairs.instReprStringPhenomenonData : Repr StringPhenomenonData

Equations

• Features.MinimalPairs.instReprStringPhenomenonData = { reprPrec := Features.MinimalPairs.instReprStringPhenomenonData.repr }

def Features.MinimalPairs.instReprStringPhenomenonData.repr : StringPhenomenonData → ℕ → Std.Format

Equations

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