Documentation

Linglib.Studies.Bobaljik2012

Bobaljik (2012): Universals in Comparative Morphology #

[Bob12]

Overview #

[Bob12] surveys comparative and superlative morphology across languages and derives several cross-linguistic generalizations from the containment hypothesis: the superlative structurally contains the comparative ([[[ADJ] CMPR] SPRL]).

Key Generalizations #

  1. CSG (Comparative-Superlative Generalizations): two directions. CSG1: a suppletive comparative forces a suppletive superlative (*ABA excluded); CSG2: a suppletive superlative forces a suppletive comparative (*AAB excluded). Attested: AAA, ABB, ABC. Unattested: *ABA, *AAB. The derivations are asymmetric — CSG1 needs containment

    • Elsewhere + Antihomophony, CSG2 additionally portmanteau exponence, adjacency, and the markedness condition (202) — in the formalization adjacency's AAB-blocking role is absorbed into Grounded; see Morphology/Containment/Vocabulary.lean. Both are instantiated on the book's worked vocabularies below. Scope: relative superlatives only, not absolute superlatives / elatives (p. 2, p. 28). The CSG ranges over synthetic grades; among periphrastic superlatives it holds exactly of those embedding the comparative form (p. 30, p. 78).
  2. SSG (Synthetic Superlative Generalization): No language has a morphological (synthetic) superlative without also having a morphological comparative.

  3. RSG (Root Suppletion Generalization): Root suppletion is limited to synthetic (morphological) comparatives. No language has a periphrastic comparative with a suppletive root (*good → more bett).

  4. Lesslessness: No language has a synthetic comparative of inferiority. "Less tall" is always periphrastic.

English and Latin Fragment Verification #

This module verifies the CSG, SSG, RSG, and lesslessness against both the English adjective fragment (Fragments/English/Modifiers/ Adjectives.lean) and the Latin adjective fragment (Fragments/Latin/ Adjectives.lean), using the suppletion field on each entry to encode empirically observed root-class patterns.

Scale-Generation Substrate #

Connects morphological infrastructure (Morphology) to scalar- alternative infrastructure (Semantics/Alternatives/HornScale.lean), enabling automatic generation of the alternatives needed for scalar implicature computation. Lives in this study file because it is the sole consumer. [Hor72] [Ken07]

A morphologically-derived Horn scale.

  • positive : String
  • comparative : String
  • superlative : String
Instances For
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      Equations
      Instances For

        Convert a MorphScale to a HornScale String, weakest-to-strongest.

        Equations
        Instances For

          Extract a degree scale from a stem's paradigm.

          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            def Bobaljik2012.ScaleFromParadigm.morphologicalAlternatives {σ : Type} (stem : Morphology.Stem σ) (form : String) :
            List String

            Paradigm-mates as scalar alternatives, scale order preserved.

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

              Pattern Verification (English) #

              CSG Verification (English) #

              CSG: All English adjective entries satisfy contiguity (no *ABA violations).

              CSG Part I applied to English data: "good" and "bad" have suppletive comparatives, so by csg_part1 their superlatives must be suppletive too — and they are.

              CSG Part II on the English data: if the superlative is suppletive, the comparative is too. Data-level check; the engine derivation (Morphology.Containment.csg2) is instantiated at welshAAB_blocked below.

              Attestedness Verification (English) #

              All English root suppletion patterns satisfy both contiguity (no *ABA — csg_part1) and the terminal-adjacent plateau (CMPR cell = SPRL cell — Morphology.Containment.realize_const_of_terminal_adjacent). The conjunction restricts root patterns to AAA / ABB. Stated as the explicit conjunction so the underlying derivation is visible at use site, rather than packaged as a stipulated isAttested field.

              SSG (Synthetic Superlative Generalization) #

              SSG ([Bob12]): If an entry has a synthetic superlative form, it also has a synthetic comparative form. No English adjective has -est without -er.

              RSG (Root Suppletion Generalization) #

              The comparative form is synthetic (a single morphological word, not periphrastic "more X"), detected as the absence of a space in the form string. Structural counterpart: Morphology.Containment.Synthesis (see the worked vocabularies below).

              Equations
              Instances For

                RSG ([Bob12]): Root suppletion is limited to synthetic comparatives. If an entry has a suppletive root (CMPR differs from POS), then its comparative form is synthetic (not periphrastic).

                This rules out patterns like *good → more bett: a language cannot have a periphrastic comparative with a suppletive root.

                Verified: "good" → "better" (synthetic), "bad" → "worse" (synthetic). Contrast: "expensive" → "more expensive" (periphrastic, but non-suppletive root — AAA, not ABB).

                Lesslessness #

                Lesslessness ([Bob12] (5), ch. 7): no language has a synthetic comparative of inferiority — "less tall" is always periphrastic. The book's empirically strongest generalization but its least derived: the proposed account (CMPR plus a polarity-reversal head, with synthetic spellout blocked by the polar antonym's comparative) is a sketch the book itself flags as incomplete, so it is recorded as data-level prose, not as a theorem of the ch. 2–5 axioms.

                Fragment Cross-Check #

                The fragment records suppletive forms for "good" and "bad".

                The suppletion field agrees with the surface forms: "good" → "better" → "best" is ABB (suppletive root shared across CMPR and SPRL).

                Cross-Linguistic Pattern Inventory #

                The three attested degree-suppletive patterns.

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

                  All attested patterns are contiguous.

                  The unattested *ABA pattern is not contiguous.

                  Cross-Linguistic Verification (Latin) #

                  Latin CSG: All Latin adjective entries satisfy contiguity.

                  Latin exhibits all three attested patterns (AAA, ABB, ABC), confirming the cross-linguistic pattern inventory against a language with richer suppletion than English.

                  The Realizational Derivation: the Book's Worked Vocabularies #

                  The engine of Morphology/Containment/Vocabulary.lean run on the vocabularies [Bob12] actually writes down. ExponenceRules record exponent, exponed span, and conditioning context; realize runs Elsewhere insertion; degreeShape reads the root-suppletion class off the realized cells. The hypotheses of the engine's theorems (csg2, exists_portmanteau_of_ne) are decide-checked on each vocabulary, and the two unattested shapes are exhibited as violations of exactly one condition each: AAB violates Grounded ((202)), surface ABA violates Antihomophonous ((44)).

                  Czech BAD ([Bob12] (39)): root allomorph hor- conditioned by CMPR, elsewhere špatn-.

                  Equations
                  • Bobaljik2012.czechBad = [{ exponent := "špatn", spans := 0, context := none }, { exponent := "hor", spans := 0, context := some 1 }]
                  Instances For
                    theorem Bobaljik2012.czech_bad_realize :
                    Morphology.Containment.realize czechBad = ![some "špatn", some "hor", some "hor"]

                    Elsewhere insertion realizes špatn-ý, hor-ší, nej-hor-ší: ABB.

                    English GOOD ([Bob12] (203)): bett- / __]CMPR], elsewhere good. Terminal and adjacent, so the plateau applies.

                    Equations
                    Instances For

                      English BAD ([Bob12] (194)): worse as a √ROOT+CMPR portmanteau, elsewhere bad. ABB arises on the portmanteau route too — worse simply realizes the whole comparative node.

                      Equations
                      Instances For

                        Welsh GOOD ([Bob12] (198)): gor- / __]SPRL] and gwell, both √ROOT+CMPR portmanteaux, elsewhere da.

                        Equations
                        • Bobaljik2012.welshGood = [{ exponent := "da", spans := 0, context := none }, { exponent := "gwell", spans := 1, context := none }, { exponent := "gor", spans := 1, context := some 2 }]
                        Instances For

                          Latin GOOD ([Bob12] (204)): opt- a √ROOT+CMPR portmanteau conditioned by SPRL, mel- a root allomorph conditioned by CMPR, elsewhere bon. The span structure also explains *opt-ior-imus: opt- swallows the CMPR cell, so regular -ior has nothing to realize.

                          Equations
                          • Bobaljik2012.latinBonus = [{ exponent := "bon", spans := 0, context := none }, { exponent := "mel", spans := 0, context := some 1 }, { exponent := "opt", spans := 1, context := some 2 }]
                          Instances For
                            theorem Bobaljik2012.latin_bonus_realize :
                            Morphology.Containment.realize latinBonus = ![some "bon", some "mel", some "opt"]

                            bon-us, mel-ior, opt-imus: the engine derives the book's star ABC paradigm.

                            Latin is not terminal — and must not be: the terminal-adjacent plateau (realize_const_of_terminal_adjacent) would force CMPR = SPRL, wrongly excluding ABC. The portmanteau rule is what escapes it.

                            theorem Bobaljik2012.latin_superlative_portmanteau :
                            Morphology.Containment.winner latinBonus 2 = some { exponent := "opt", spans := 1, context := some 2 }

                            Latin's superlative winner computed concretely: the portmanteau opt- — the ABC-requires-portmanteau consequence ([Bob12] §5.3.1, generalized there as (199)).

                            The engine theorem applied: since Latin's comparative and superlative cells diverge, exists_portmanteau_of_ne forces a portmanteau winner at the superlative.

                            The nanosyntax lexicon for Latin GOOD, in the [Cah09] method (the degree application is later literature; see Morphology/Containment/Superset.lean): context-free entries storing successively larger constituents, competing under the Superset Principle.

                            Equations
                            • Bobaljik2012.latinBonusNS = [{ exponent := "bon", spans := 0, context := none }, { exponent := "mel", spans := 1, context := none }, { exponent := "opt", spans := 2, context := none }]
                            Instances For
                              theorem Bobaljik2012.latin_ns_spellout :
                              Morphology.Containment.spellout latinBonusNS = ![some "bon", some "mel", some "opt"]

                              Superset spellout derives the same ABC paradigm with no contextual apparatus — the entries are portmanteaus by constituent size alone, doing the work of DM's context-restricted ch. 5 machinery.

                              Cross-framework agreement on the data: the nanosyntax lexicon and the DM vocabulary (204) realize the same paradigm cell for cell (the concrete face of Morphology.Containment.spelloutGenerable_iff_generable).

                              The hypothetical AAB vocabulary ([Bob12] (201)): gor- restricted to the superlative with no comparative-level counterpart. It generates the unattested AAB shape (*da – da-ch – gor-au)...

                              Equations
                              • Bobaljik2012.welshAAB = [{ exponent := "da", spans := 0, context := none }, { exponent := "gor", spans := 1, context := some 2 }]
                              Instances For

                                ...and is excluded by exactly one condition: its threshold set skips the comparative grade, violating Grounded ((202)).

                                The engine theorem csg2 applied: no vocabulary realizing the AAB cells can satisfy both Antihomophony and Grounded — the AAB realization itself refutes the conjunction.

                                The homophonous-ABC loophole ([Bob12] (44) discussion): without Antihomophony, surface ABA is generable — a superlative root allomorph accidentally homophonous with the positive.

                                Equations
                                • Bobaljik2012.fakeAba = [{ exponent := "A", spans := 0, context := none }, { exponent := "B", spans := 0, context := some 1 }, { exponent := "A", spans := 0, context := some 2 }]
                                Instances For

                                  Synthesis: the Merger layer on the worked vocabularies #

                                  The structural synthetic/analytic notion (Morphology.Containment.Synthesis) on English good: the word merges through the superlative (wordTop = 2), and since the word-internal realization shows distinct root forms at the positive and comparative grades, rsg certifies the comparative as synthetic — the structural counterpart of english_rsg's string-level check.

                                  Scale Generation from Degree Paradigms #

                                  ScaleFromParadigm (the Scale-Generation Substrate section above) derives Horn scales from degree paradigms: a stem with comparative + superlative rules yields a 3-point scale [positive, comparative, superlative]. The tests below verify the extractor on the English adjective fragment.

                                  Gradable adjectives produce a scale; non-gradables do not.

                                  Regular paradigm yields the expected 3-point scale.

                                  Suppletive paradigm yields the irregular forms in scale position.

                                  theorem Bobaljik2012.expensive_scale_members :
                                  Option.map (fun (x : ScaleFromParadigm.MorphScale) => x.toHornScale.members) (ScaleFromParadigm.adjectiveScale Bobaljik2012.expensiveStem✝) = some ["expensive", "more expensive", "most expensive"]

                                  Periphrastic paradigm yields multi-word scale members.

                                  Morphological Alternatives #

                                  morphologicalAlternatives stem form returns paradigm-mates of form preserving scale order — the input shape scalar-implicature reasoning expects. Tests confirm filter-by-equality semantics across the three scale positions, plus the empty-list result for non-gradable stems.