Documentation

Linglib.Phenomena.Morphology.Studies.Bybee1985

Bybee (1985): Morphology — A Study of the Relation Between Meaning and Form #

@cite{bybee-1985}

@cite{bybee-1985} (Typological Studies in Language 9, John Benjamins) reports a 50-language sample (Perkins 1980 stratified probability sample) and tests three predictions of the relevance hypothesis: morpheme categories whose meaning is more directly relevant to the verb stem (a) occur more frequently as inflectional categories cross-linguistically, (b) occur closer to the stem in suffixal morphology, and (c) exhibit greater fusion with the stem.

This study file formalizes the cross-linguistic data underlying claims (a) and (b) — Ch 2 §5 (frequency, Figs 1+2) and Ch 2 §6 (morpheme order, pp. 34-35) — and connects them to the substrate's MorphCategory.peripherality ranking. Claim (c) (fusion) is qualitative in Ch 2 §7 and not formalized here.

What this file does NOT cover. Ch 1 fusion/allomorphy/grammatical meaning, Ch 3 paradigm organization (basic-derived, local markedness), Ch 4 lexical-derivational-inflectional CONTINUUM (which contradicts the substrate's discrete MorphStatus enum — flagged for future work), Part II aspect/tense/mood detail (Ch 6-9). See Phenomena/Morphology/Studies/ HahnDegenFutrell2021.lean for a downstream consumer of the §6 ordering data via RespectsRelevanceHierarchy. Ch 5's dynamic network model substrate now lives in Theories/Morphology/UsageBased/Network.lean; §5 below uses it on an English sing/sang/sung instance.

Bybee discusses six core verbal-inflectional categories ordered by relevance to the verb stem (most relevant first): valence, voice, aspect, tense, mood, agreement. Number, person, and gender are sub-types of agreement; object-person is reported separately in the survey data.

source
inductive Phenomena.Morphology.Studies.Bybee1985.BybeeCategory :
Type

The verbal-inflectional categories Bybee surveys in Ch 2.

The constructors are listed in Bybee's relevance order (most relevant to verb stem first; cf. Ch 2 §3 p. 20 which orders "valence, voice, aspect, tense, mood and agreement"). Object-person, number-as-agreement, and gender are reported separately in Figs 1+2 (p. 30-31).

Instances For
    source
    @[implicit_reducible]
    instance Phenomena.Morphology.Studies.Bybee1985.instDecidableEqBybeeCategory :
    DecidableEq BybeeCategory
    Equations
    source
    @[implicit_reducible]
    instance Phenomena.Morphology.Studies.Bybee1985.instReprBybeeCategory :
    Repr BybeeCategory
    Equations
    source
    def Phenomena.Morphology.Studies.Bybee1985.instReprBybeeCategory.repr :
    BybeeCategory → ℕ → Std.Format
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      source
      @[implicit_reducible]
      instance Phenomena.Morphology.Studies.Bybee1985.instFintypeBybeeCategory :
      Fintype BybeeCategory
      Equations
      • One or more equations did not get rendered due to their size.

      Figure 1 (p. 30) reports the percentage of the 50-language sample with inflectional expression of each category. Figure 2 (p. 31) reports the same categories counting both inflectional and derivational affixes. The raw counts (percentage × 50 / 100) are integers because the sample is exactly 50 languages.

      source
      def Phenomena.Morphology.Studies.Bybee1985.inflectionalCount50 :
      BybeeCategory → ℕ

      Number of languages (out of 50) that have inflectional expression of the given category. Verified directly against @cite{bybee-1985} Fig 1 (p. 30). The raw counts equal 0.01 × percentage × 50.

      Equations
      Instances For
        source
        def Phenomena.Morphology.Studies.Bybee1985.derivOrInflCount50 :
        BybeeCategory → ℕ

        Number of languages (out of 50) with inflectional or derivational expression of the category. Fig 2 (p. 31). Valence's 90% reflects the near-universality of valence-changing morphology (Ch 2 §5 p. 29) when derivational expression is included; only Haitian, Karankawa, Navaho, Serbo-Croatian, and Vietnamese show no such morphology in Bybee's sources.

        Equations
        Instances For
          source
          theorem Phenomena.Morphology.Studies.Bybee1985.valence_highest_when_derivOrInfl (c : BybeeCategory) :
          derivOrInflCount50 c ≤ derivOrInflCount50 BybeeCategory.valence

          Bybee's first prediction (Ch 2 §5 p. 29-30): when both inflectional and derivational expression are counted, valence shows the highest cross- linguistic frequency, reflecting its near-universality as a derivational process even in languages with sparse verbal morphology.

          source
          theorem Phenomena.Morphology.Studies.Bybee1985.mood_highest_when_inflectional (c : BybeeCategory) :
          inflectionalCount50 c ≤ inflectionalCount50 BybeeCategory.mood

          Among purely inflectional categories, mood shows the highest frequency (Fig 1, p. 30: 68%). The valence drop from 90% (deriv+infl) to 6% (infl only) is Bybee's central observation that valence-changing morphology is "almost always derivational" (Ch 2 §5 p. 30).

          source
          theorem Phenomena.Morphology.Studies.Bybee1985.gender_lowest_when_derivOrInfl (c : BybeeCategory) :
          derivOrInflCount50 BybeeCategory.genderAgr ≤ derivOrInflCount50 c

          In the deriv+infl survey (Fig 2), gender agreement is the lowest- frequency category — consistent with Bybee's claim (Ch 2 §3 p. 23) that gender agreement has least relevance to the verb. (In Fig 1 inflection- only, valence drops to 6% — below gender's 16% — because valence is near-totally suppressed as inflection while remaining 90% as derivation. The Fig 2 ranking is the relevance-faithful one.)

          Bybee's second prediction (Ch 2 §6 p. 33): the morpheme-order corollary. "The most relevant to occur closest to the verb stem, and the least relevant to occur at the greatest distance from the verb stem." Tested on the four most frequent categories — aspect, tense, mood, person — by counting, for each pair, in how many languages one is closer to the stem than the other.

          Cases were excluded when (a) the morphemes were portmanteau, (b) they appeared on different sides of the stem (unless one was adjacent and the other was not), (c) they were mutually exclusive in the same slot, or (d) one was expressed via stem modification (counted as closer).

          source
          structure Phenomena.Morphology.Studies.Bybee1985.OrderPair :
          Type

          A morpheme-order pair from Ch 2 §6. The fields record:

          • closer — the category Bybee predicts to be closer to the stem
          • further — the category Bybee predicts to be further from the stem
          • closerCount — # languages where closer is closer to stem (predicted)
          • furtherCount — # languages where further is closer to stem (counter-ex)
          • total — total # languages where the pair is testable (Ch 2 §6 p. 34-35)
          Instances For
            source
            def Phenomena.Morphology.Studies.Bybee1985.instReprOrderPair.repr :
            OrderPair → ℕ → Std.Format
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              source
              @[implicit_reducible]
              instance Phenomena.Morphology.Studies.Bybee1985.instReprOrderPair :
              Repr OrderPair
              Equations
              source
              def Phenomena.Morphology.Studies.Bybee1985.orderPairs :
              List OrderPair

              The six pairs Bybee tests in Ch 2 §6 (pp. 34-35). All counts verified directly against the book.

              Equations
              • One or more equations did not get rendered due to their size.
              Instances For
                source
                theorem Phenomena.Morphology.Studies.Bybee1985.aspect_categorical_against_tense_and_mood :
                Option.map (fun (x : OrderPair) => x.furtherCount) orderPairs.head? = some 0 ∧ Option.map (fun (x : OrderPair) => x.furtherCount) orderPairs[1]? = some 0

                Aspect vs. tense and aspect vs. mood are categorical: zero counter- examples in the entire 50-language sample. Bybee's summary (p. 35): "the strongest differences are found between aspect and the other categories, and between tense and the other categories, where there are almost no counter-examples to the predicted ordering."

                source
                theorem Phenomena.Morphology.Studies.Bybee1985.mood_person_ordering_is_freer (p : OrderPair) :
                p ∈ orderPairs → p.further = BybeeCategory.personAgr ∧ p.closer = BybeeCategory.mood ∨ 10 * p.furtherCount ≤ p.total

                Mood-person ordering is genuinely freer: the mood-person pair is the only one in the survey where the predicted-counter / total ratio exceeds 1/10.

                source
                theorem Phenomena.Morphology.Studies.Bybee1985.predicted_outnumbers_counter :
                (orderPairs.all fun (p : OrderPair) => decide (p.furtherCount < p.closerCount)) = true

                Across all six tested pairs, the predicted direction outnumbers the counter-direction in every case (Ch 2 §6 summary, p. 35: "striking confirmation of the hierarchical ordering of aspect, tense, mood and person").

                source
                theorem Phenomena.Morphology.Studies.Bybee1985.ch2_section6_aggregate_counts :
                (List.map (fun (x : OrderPair) => x.total) orderPairs).sum = 125 ∧ (List.map (fun (x : OrderPair) => x.closerCount) orderPairs).sum = 59 ∧ (List.map (fun (x : OrderPair) => x.furtherCount) orderPairs).sum = 8

                Total testable observations across all six pairs: 125 language-pair observations contributed to the hierarchy test. The predicted direction holds in 59 of them; the counter-direction in 8; the remaining 58 are non-relevant (portmanteau, opposite-side, mutually- exclusive, or stem-modification cases per Bybee's exclusion criteria on p. 34).

                Core/Lexical/MorphRule.lean::MorphCategory.peripherality numerically encodes Bybee's relevance hierarchy: lower number = closer to stem = more relevant. The substrate is faithful to Bybee Ch 2 §3 (p. 20-23) for the six core categories Bybee discusses (valence < voice < aspect < tense < mood < agreement). The substrate adds extensions (derivation, degree, negation, nonfinite) that are not part of Bybee's verbal hierarchy — those are flagged in MorphRule.lean's docstring as linglib extensions.

                source
                def Phenomena.Morphology.Studies.Bybee1985.toMorphCategory :
                BybeeCategory → Core.Morphology.MorphCategory

                Map Bybee's category enum to the substrate's MorphCategory. Object agreement and number-as-agreement collapse to the substrate's .agreement and .number respectively.

                Equations
                Instances For
                  source
                  theorem Phenomena.Morphology.Studies.Bybee1985.substrate_matches_bybee_hierarchy :
                  List.Pairwise (fun (x1 x2 : ℕ) => x1 < x2) (List.map (fun (c : BybeeCategory) => (toMorphCategory c).peripherality) [BybeeCategory.valence, BybeeCategory.voice, BybeeCategory.aspect, BybeeCategory.tense, BybeeCategory.mood, BybeeCategory.personAgr])

                  The substrate's peripherality ordering is strictly monotonic on the six categories Bybee discusses in Ch 2 §3 (excluding the agreement sub-types which collapse to a single rank). In other words, the substrate faithfully reproduces Bybee's high-relevance-end ordering: valence < voice < aspect < tense < mood < agreement.

                  source
                  theorem Phenomena.Morphology.Studies.Bybee1985.orderPairs_consistent_with_substrate :
                  (orderPairs.all fun (p : OrderPair) => decide ((toMorphCategory p.closer).peripherality < (toMorphCategory p.further).peripherality) || decide ((toMorphCategory p.closer).peripherality = (toMorphCategory p.further).peripherality)) = true

                  The §6 morpheme-order data is exactly what RespectsRelevanceHierarchy predicts: when Bybee's predicted-closer category has lower peripherality than the predicted-further one, the substrate predicate agrees with the empirical-majority direction.

                  Bybee Ch 5 §8 (p. 124) illustrates the network architecture with the Spanish dormir paradigm: three stem allomorphs (dwerm, dorm, durmy) connected by morphological relations across persons/numbers/ tenses. We use the English irregular verb eat/ate/eaten as a smaller instance — but rather than declaring fabricated LexicalEntry strings, we derive the network from the actual eat : VerbEntry in Fragments/English/Predicates/Verbal.lean. Per CLAUDE.md "derive don't duplicate": changing eat.formPast in the Fragment automatically updates this network; the connection between Fragment data and Bybee-network claims is true by construction.

                  Token frequencies default to 0 in the bridge — Bybee's verified per- lexeme counts (Francis & Kučera 1982) live in Theories/Morphology/UsageBased/Network.lean's strongStillStrong / strongRegularized datasets, which support the Ch 5 §6 irregularity- vs-frequency theorem.

                  source
                  def Phenomena.Morphology.Studies.Bybee1985.verbToLexicalEntries (v : Fragments.English.Predicates.Verbal.VerbEntry) :
                  List (Theories.Morphology.UsageBased.LexicalEntry Unit)

                  Extract a verb's five inflected forms as Bybee LexicalEntry instances. Token frequencies default to 0; the Fragment doesn't carry per-form frequencies.

                  Equations
                  • One or more equations did not get rendered due to their size.
                  Instances For
                    source
                    def Phenomena.Morphology.Studies.Bybee1985.verbToBybeeNetwork (v : Fragments.English.Predicates.Verbal.VerbEntry) :
                    Theories.Morphology.UsageBased.Network Unit

                    Construct the Bybee network of a verb's paradigm from its Fragment VerbEntry. All inflected forms receive pairwise semantic connections (they share the verb's meaning, modulated by tense/ aspect/agreement). Pairwise phonological connections approximate Bybee's "shared phonological skeleton" relation; a real similarity metric would gate this more selectively (e.g. eat/ate share /e_t/ but not /eating/ proper).

                    Equations
                    • One or more equations did not get rendered due to their size.
                    Instances For
                      source
                      def Phenomena.Morphology.Studies.Bybee1985.eatNetwork :
                      Theories.Morphology.UsageBased.Network Unit

                      The Bybee network of the English irregular verb eat — derived from the Fragment, NOT stipulated.

                      Equations
                      Instances For
                        source
                        theorem Phenomena.Morphology.Studies.Bybee1985.eat_past_in_network :
                        (eatNetwork.entries.any fun (x : Theories.Morphology.UsageBased.LexicalEntry Unit) => x.form == "ate") = true

                        Sanity check: eat's past form is ate (verified via the network's presence test, which derives from the Fragment's eat.formPast field). If eat.formPast were changed in Verbal.lean from "ate" to anything else, this theorem would fail to type-check.

                        source
                        theorem Phenomena.Morphology.Studies.Bybee1985.ch5_section6_anchor_lives_in_substrate :
                        (List.map (fun (x : Theories.Morphology.UsageBased.LexicalEntry Unit) => x.tokenFreq) Theories.Morphology.UsageBased.strongStillStrong).sum * Theories.Morphology.UsageBased.strongRegularized.length > (List.map (fun (x : Theories.Morphology.UsageBased.LexicalEntry Unit) => x.tokenFreq) Theories.Morphology.UsageBased.strongRegularized).sum * Theories.Morphology.UsageBased.strongStillStrong.length

                        The empirical anchor for Ch 5 §6's central claim — that high- frequency Strong Verbs maintained their irregularity while low- frequency ones regularized — lives in the substrate file as strong_verbs_higher_frequency_than_regularized (Bybee p. 120, verified Francis & Kučera 1982 counts across Sweet's Anglo Saxon Primer Class I/II/VII). Re-stated here so the study file's coverage is self-documenting.