Documentation

Linglib.Phenomena.Morphology.Studies.Link1983

Morphological Composition: Phenomena #

@cite{link-1983}

Empirical tests for the Core.Morphology pipeline, verifying that stem-based generation produces correct surface forms and features from real English lexical entries in Fragments/English/.

Coverage #

  1. Regular plurals: dog → dogs, student → students
  2. Irregular plurals: man → men, child → children, person → people
  3. Mass nouns: water, furniture — no plural paradigm
  4. Proper names: John, Mary — no plural paradigm
  5. Verb agreement: sleep → sleeps (vacuous), eat → ate (vacuous)
  6. Vacuity: all verb inflection is semantically vacuous
theorem Phenomena.Morphology.CompositionBridge.plural_changes_truth_conditions :
∃ (atomPred : BoolBool) (closurePred : (BoolBool)BoolBool) (pred : BoolBool) (x : Bool), (Core.Morphology.Number.pluralNounRule closurePred atomPred).semEffect pred x pred x

With the mereological plural rule, plural marking is NOT semantically vacuous: there exist predicates and entities for which the plural denotation differs from the base. This witnesses @cite{link-1983}'s point that singular and plural nouns denote distinct sets.

The Zwicky & Pullum CliticAffixProfile and the MorphRule framework characterize the same morphological operations from complementary angles:

The bridge theorems below verify that these independent characterizations agree on whether a morphological operation carries semantic content.

The plural -s affix is classified as an inflectional affix by Zwicky's diagnostics, AND is semantically non-vacuous in the MorphRule framework. Both analyses agree that plural marking carries semantic content.

The past tense -ed affix is classified as inflectional by Zwicky's diagnostics. In the MorphRule framework, verb tense inflection is vacuous at the word level (tense semantics is compositional, handled by Semantics.Intensional). This is NOT a disagreement: Zwicky's hasSemanticIdiosyncrasies tracks compositionality failures (like "last" ≠ "most late"), while MorphRule.isVacuous tracks whether the regular semantic contribution is non-trivial.

The @cite{bybee-1985} relevance hierarchy orders morpheme categories by semantic relevance to the verb stem. Categories with higher relevance rank (e.g., agreement = 8) are less semantically relevant — and therefore more likely to be semantically vacuous.

We verify that MorphRule.isVacuous is consistent with Bybee's predictions: agreement morphology (rank 8, least relevant) is vacuous, while number on nouns (which changes denotation via @cite{link-1983}) is not.

Verb agreement (3sg -s) has category =.agreement, which has the highest relevance rank (= least relevant to verb meaning). This predicts semantic vacuity — and indeed verbAgreement3sg has isVacuous = true.

Plural noun marking has category =.number, distinct from .agreement — it changes the noun's denotation. The full mereological rule has isVacuous = false.

The .number and .agreement categories have different relevance ranks, reflecting that noun number is semantically contentful while verb agreement is vacuous.

Per-entry regularity verification: changing a verb's isRegular flag breaks exactly one theorem.

Rule-grounding for regular verbs: stored forms match expected output. Concrete string checks close by rfl; changing a verb's stem or form fields breaks exactly the relevant theorem.

Batch consistency: for every regular verb, the stored forms match the rule-computed forms. A single native_decide covers all entries.