Documentation

Linglib.Theories.Syntax.ConstructionGrammar.Studies.FillmoreKayOConnor1988

@cite{fillmore-kay-oconnor-1988}: Let Alone #

Formalization of "Regularity and Idiomaticity in Grammatical Constructions: The Case of Let Alone" (Language 64(3):501–538).

This foundational Construction Grammar paper argues that let alone is a formal idiom: a productive syntactic pattern with non-compositional semantics and specific pragmatic constraints. The key contributions:

  1. Idiom typology: encoding vs decoding, grammatical vs extragrammatical, substantive vs formal (§1.1–1.2)
  2. Scalar model: n-dimensional argument space with a monotonicity constraint on propositional functions (Appendix, Definition A3)
  3. Let alone construction: form F ⟨X A Y let alone B⟩ requires A and B to be points on a presupposed scale, with F'⟨X A Y⟩ entailing F'⟨X B Y⟩
  4. Pragmatic function: resolves conflict between Gricean Quantity (informativeness — the A clause) and Relevance (the B clause)

Section 1: Idiom Typology (§1.1–1.2) #

Fillmore et al.'s classification cross-cuts two dimensions:

Encoding vs decoding idioms (§1.1.1, @cite{makkai-1972}).

A decoding idiom cannot be interpreted without prior learning ("kick the bucket"). An encoding idiom can be understood but its conventional status must be learned ("answer the door").

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

      Grammatical vs extragrammatical idioms (§1.1.2).

      Grammatical idioms have words filling proper grammatical slots ("kick the bucket"). Extragrammatical idioms have anomalous structure ("first off", "by and large", "so far so good").

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

          Substantive vs formal idioms (§1.1.3).

          Substantive (lexically filled) idioms have fixed lexical content. Formal idioms are syntactic patterns dedicated to semantic/pragmatic purposes not knowable from form alone.

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

              Combined idiom classification.

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

                  Familiar-pieces typology (§1.2): how familiar are the pieces and their mode of combination?

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

                      Section 2: Scalar Models (§2.3.2, Appendix) #

                      The formal backbone of the paper: an n-dimensional scalar model with a monotonicity constraint on propositional functions.

                      Definition A3: (S, T, D^x, P) is a SCALAR MODEL iff, for distinct d_i, d_j in D^x, P(d_j) entails P(d_i) just in case d_i is LOWER than d_j.

                      Where "lower" means: d_i is lower than d_j iff no coordinate of d_i has a higher value than the corresponding coordinate of d_j, and at least one coordinate of d_i has a lower value (Definition A2).

                      An argument point in the n-dimensional argument space D^x. In the paper's example: (Apotheosis, English) is an argument point in the 2D space of linguists × languages.

                      • coordinates : List α

                        Coordinates, one per dimension

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

                            A scalar model (Definition A3 from the Appendix).

                            Given argument space D^x and propositional function P, the scalar model constrains P so that lower argument points yield weaker (entailed) propositions.

                            We use Bool for decidable propositions over states S.

                            • points : List (ArgumentPoint α)

                              Argument points (elements of D^x)

                            • propFn : ArgumentPoint αSBool

                              Propositional function: argument point → proposition over states

                            • dimLe : ααBool

                              Ordering on individual dimension values

                            Instances For

                              An argument point d_i is LOWER than d_j (Definition A2): no coordinate of d_i exceeds the corresponding coordinate of d_j, and at least one coordinate of d_i is strictly lower.

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

                                Scalar entailment: P(d_j) entails P(d_i) iff {s | P(d_j)(s)} ⊆ {s | P(d_i)(s)}.

                                In a valid scalar model, this holds exactly when d_i is lower than d_j.

                                Equations
                                Instances For

                                  Informativeness/strength (Definition A5): p is MORE INFORMATIVE (STRONGER) than q relative to a scalar model iff p entails q in SM and q does not entail p in SM.

                                  Equations
                                  Instances For

                                    Definition A3 validity: a scalar model satisfies the monotonicity constraint iff for all distinct argument points d_i, d_j in D^x, P(d_j) entails P(d_i) exactly when d_i is lower than d_j.

                                    We check the forward direction (lower → entails) for all point pairs. This is the computable check used by native_decide.

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

                                      Section 3: The Let Alone Construction (§2.1–2.4) #

                                      The let alone construction as a Construction.

                                      Form: F ⟨X A Y let alone B⟩

                                      • F = negative polarity operator (negation, doubt, barely, etc.)
                                      • X, Y = shared non-focused material
                                      • A = first focused element (in the stronger, full clause)
                                      • B = second focused element (in the weaker, reduced clause)
                                      • A and B must be points on a presupposed scale
                                      Equations
                                      • One or more equations did not get rendered due to their size.
                                      Instances For

                                        Semantic conditions on let alone sentences (p.528).

                                        For a let alone sentence to be well-formed:

                                        1. The full clause and reduced clause are propositions from the same scalar model
                                        2. The two propositions are of the same polarity
                                        3. The full clause proposition is stronger (more informative) than the reduced clause
                                        Instances For

                                          The let alone family: related conjunctions with similar scalar semantics (p.533). All presuppose a scalar model relating their two conjuncts. They differ in clause ordering (stronger-first vs weaker-first).

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

                                              Section 4: NPI status (§2.2.4) #

                                              Let alone is a negative polarity item, but with nuances:

                                              NPI trigger types that license let alone (§2.2.4, exx.62-70).

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

                                                  Section 5: Construction definitions for the constructicon #

                                                  The X-er the Y-er comparative correlative (§1.2.1, exx.1-2).

                                                  A formal idiom with unfamiliar pieces unfamiliarly arranged. The definite article "the" is unique to this construction (p.507) — a fixed element — and the two-part structure has no parallel elsewhere in English. The open comparative phrases make it partially open (mix of fixed "the" and open comparatives).

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

                                                    The Incredulity Response construction (§1.1.4, ex.14h).

                                                    "Him be a doctor?" — non-nominative subject + bare stem verb, expressing incredulity. A formal idiom: the pattern (NP[acc] VP[bare]) is productive and dedicated to a rhetorical/evaluative pragmatic function not derivable from its component meanings.

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

                                                      Inheritance: let alone inherits from coordinating conjunction but overrides several properties.

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

                                                        Section 6: 1D Scalar Model — Military Ranks (§2.1, ex.21) #

                                                        The running example: ⟨second lieutenant,..., colonel, general⟩.

                                                        Military ranks from the paper's running example.

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

                                                            Rank ordering (lower index = lower rank).

                                                            Equations
                                                            Instances For

                                                              States: whether a person achieved each rank.

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

                                                                  The military rank scalar model.

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

                                                                    The rank scalar model is one-dimensional: every argument point has exactly one coordinate. The paper argues (fn.16, p.535) that some examples require ≥2 dimensions.

                                                                    Scalar entailment: "He made general" entails "He made colonel". This is the core semantic condition on let alone (p.523, 528): the stronger (A) proposition entails the weaker (B) proposition.

                                                                    Scalar entailment: "He made colonel" entails "He made lieutenant".

                                                                    The reverse does NOT hold: "He made colonel" does not entail "He made general". This is why "He didn't make colonel, let alone general" is felicitous — general is STRONGER (p.528).

                                                                    Making general is STRONGER than making colonel: the extension of "made general" is a proper subset of "made colonel" (Definition A5).

                                                                    Second lieutenant is the LOWEST point: no other rank is lower. This explains the anomaly in ex.107: "let alone a second lieutenant" is anomalous because you cannot scalar-entail the negation of the lowest point from the negation of a non-lowest point (p.526).

                                                                    The rank scalar model validates FKO's ex.21: "He didn't make colonel, let alone general."

                                                                    Under negation, the scalar direction reverses: "not make colonel" entails "not make general" (if you can't do the easier thing, you can't do the harder thing).

                                                                    Section 7: 2D Scalar Model — Linguists × Languages (§2.3.2, Tables 1–4) #

                                                                    The paper's most carefully developed example: four professors (Apotheosis, Brilliant, Competent, Dimm) and four languages (English, French, Greek, Hittite), ordered by erudition and accessibility respectively.

                                                                    The propositional function P: "professor X can read language L" is monotone: if X is more erudite and L is more accessible, P is more likely to be true.

                                                                    This 2D model is the basis for the Appendix definitions (A1–A5).

                                                                    Linguists ordered by erudition (most → least).

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

                                                                        Languages ordered by accessibility (most → least).

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

                                                                            Dimension value: either a linguist or a language. The argument space D^x = Linguist × Lang, encoded as 2-element coordinate lists of LingLangVal.

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

                                                                                  Dimension ordering (≤) for the linguist/language scalar model.

                                                                                  From Definition A2 (p.536) and the worked example on p.537: d_i is LOWER than d_j when d_i has coordinatewise ≤ values with at least one strict. A LOWER argument point yields a WEAKER proposition (true in more states).

                                                                                  • Linguist: Apotheosis ≤ Brilliant ≤ Competent ≤ Dimm (more erudite = LOWER = weaker — "Apotheosis reads L" is easiest to satisfy)
                                                                                  • Language: English ≤ French ≤ Greek ≤ Hittite (more accessible = LOWER = weaker — "X reads English" is easiest to satisfy)

                                                                                  The paper confirms (p.537): "(B, E) is lower than (B, G)" — Brilliant with English is lower than Brilliant with Greek. Cross-type comparisons return false (dimensions are independent).

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

                                                                                    States of affairs for the linguist/language model. Each state specifies which (linguist, language) pairs satisfy "X can read L". We use a few representative states from Table 2 in the paper (p.527).

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

                                                                                        Convenience constructor for 2D argument points.

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

                                                                                          The linguist/language scalar model from §2.3.2 (Tables 1–4, p.526–527).

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

                                                                                            In the 2D model, "Brilliant can read Hittite" entails "Brilliant can read English" — Hittite is less accessible, so knowing it is stronger (p.528, exx.109–112).

                                                                                            "Brilliant can read Hittite" is stronger than "Competent can read French" (Definition A5). Note: these points are INCOMPARABLE in the partial order (Brilliant < Competent on linguist, Hittite > French on language), but the entailment holds empirically — the scalar model constrains more than just comparable pairs.

                                                                                            (Brilliant, English) is lower than (Brilliant, Greek): the paper's own worked example (p.537). Same linguist, but English is more accessible (lower) than Greek.

                                                                                            (Competent, French) and (Brilliant, Hittite) are INCOMPARABLE: Competent > Brilliant on linguist but French < Hittite on language. Neither is uniformly ≤ the other (Definition A2).