Bresnan 1973: Syntax of the Comparative Clause Construction in English #
@cite{bresnan-1973}
Joan W. Bresnan. Syntax of the Comparative Clause Construction in English. Linguistic Inquiry 4(3): 275–343.
Core Contributions #
QP structure: Every comparative has an underlying QP (Det + Q) where the Q items are
much,many,little,few,enough, and the Det position hosts-er,-est,as,too,so,that, or∅. The surface formmorederives from-er+muchvia suppletion.Morphological derivation: Three ordered rules derive surface forms:
Comparative Deletion: The
than/asclause originates in the Det of QP. An obligatory deletion operation removes a clause constituent "nondistinct" from the head. The clause is then extraposed.All than-clauses are underlyingly clausal: "Phrasal" comparatives like
taller than Billderive from full clauses via maximal deletion. This contrasts with the 2026 consensus (@cite{bhatt-pancheva-2004}) that phrasal and clausal comparatives are syntactically distinct.Subdeletion: When the head is a QP (measure phrase), only a matching QP is deleted from the clause — not an AP or NP. This explains
*more than Bill tall(NP ≠ QP identity failure).Privative adjective constraint:
*more than five feet shortbecauseshortrejects definite measure phrases, so the QP identity condition in the clause cannot be satisfied.
What Is Current vs. Historical #
Current consensus: The data paradigms, the degree-head inventory (Det →
modern Deg°), the more = -er + much decomposition, the measure phrase
restriction with negative adjectives, subdeletion commensurability.
Modified but descended: QP → DegP (@cite{kennedy-1999}); comparative deletion → degree operator movement (@cite{heim-2001}); QP-AP parallelism → X-bar theory → Minimalist bare phrase structure.
Historical: The construction-specific transformations (AP Shift, QP
Raising, Much Deletion, -er Encliticizing), the "all comparatives are
clausal" claim, the so → such transformation.
Connection to Kennedy 1999 #
The existing Kennedy1999.lean cites @cite{bresnan-1973} for morphological
distribution data and phrasal/clausal examples. This file formalizes
Bresnan's own proposals: the QP structure, the derivation rules, and the
four introductory puzzles that motivated the analysis. Bridge theorems
connect the QP inventory to DegPType and verify suppletion outputs
against Fragment formComp entries.
Equations
- Phenomena.Comparison.Studies.Bresnan1973.instBEqDet.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Phenomena.Comparison.Studies.Bresnan1973.instBEqQ.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
The QP: Bresnan's degree phrase structure.
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
Map Bresnan's QP Det to the modern DegP type classification. The Det determines the comparison type; Q determines mass/count.
Equations
- Phenomena.Comparison.Studies.Bresnan1973.Det.er.toDegPType = Semantics.Degree.DegPType.comparative
- Phenomena.Comparison.Studies.Bresnan1973.Det.est.toDegPType = Semantics.Degree.DegPType.superlative
- Phenomena.Comparison.Studies.Bresnan1973.Det.as_.toDegPType = Semantics.Degree.DegPType.equative
- Phenomena.Comparison.Studies.Bresnan1973.Det.too.toDegPType = Semantics.Degree.DegPType.excessive
- Phenomena.Comparison.Studies.Bresnan1973.Det.null.toDegPType = Semantics.Degree.DegPType.sufficiency
- Phenomena.Comparison.Studies.Bresnan1973.Det.so.toDegPType = Semantics.Degree.DegPType.excessive
- Phenomena.Comparison.Studies.Bresnan1973.Det.that_.toDegPType = Semantics.Degree.DegPType.excessive
- Phenomena.Comparison.Studies.Bresnan1973.Det.any_.toDegPType = Semantics.Degree.DegPType.comparative
- Phenomena.Comparison.Studies.Bresnan1973.Det.no_.toDegPType = Semantics.Degree.DegPType.comparative
Instances For
The modern DegPType inventory is recoverable from Bresnan's Det.
Suppletion: the surface form resulting from -er/-est Encliticizing
onto Q. Returns none for regular (non-suppletive) combinations.
Rule (7): -er much → more, -er many → more, -er little → less Rule (223): -est much → most, -est many → most, etc.
Equations
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.er, q := Phenomena.Comparison.Studies.Bresnan1973.Q.much } = some "more"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.er, q := Phenomena.Comparison.Studies.Bresnan1973.Q.many } = some "more"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.er, q := Phenomena.Comparison.Studies.Bresnan1973.Q.little } = some "less"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.er, q := Phenomena.Comparison.Studies.Bresnan1973.Q.few } = some "fewer"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.est, q := Phenomena.Comparison.Studies.Bresnan1973.Q.much } = some "most"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.est, q := Phenomena.Comparison.Studies.Bresnan1973.Q.many } = some "most"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.est, q := Phenomena.Comparison.Studies.Bresnan1973.Q.little } = some "least"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion { det := Phenomena.Comparison.Studies.Bresnan1973.Det.est, q := Phenomena.Comparison.Studies.Bresnan1973.Q.few } = some "fewest"
- Phenomena.Comparison.Studies.Bresnan1973.suppletion x✝ = none
Instances For
most derives from -est + much.
Equations
Instances For
Much Deletion: much → ∅ / [... ___ A]_AP.
Applies only to Qs that can modify adjectives, and only when an
adjective or adverb immediately follows.
Equations
- Phenomena.Comparison.Studies.Bresnan1973.muchDeletionApplies q adjFollows = (q.canModifyAdjective && adjFollows)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Puzzle (A): "I've never seen a man taller than my father/mother."
(i) and (ii) are roughly synonymous; (iii) and (iv) are not. (i) "a man taller than my father" — head: AP [x much tall] (ii) "a taller man than my father" — head: AP [x much tall] (iii) "a man taller than my mother" — head: AP [x much tall] (iv) "a taller man than my mother" — head: NP [a man x much tall]
In (iv), the head is the entire predicative NP. The deleted constituent in the clause must match this NP — but "my mother is [x much tall] a man" implies my mother is a man. Hence the anomaly.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Puzzle (B): "Jack eats caviar more than he eats mush / sleeps."
(a,c) are grammatical; (d) is not. The deleted constituent is an adverbial QP modifier of the VP.
In (d), *Jack eats more caviar than he sleeps is bad because
more is a partitive QP embedded in the direct object NP. The
matching partitive QP in the clause requires an NP — but sleep
is intransitive, so no matching partitive is available.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Puzzle (C): "I am more angry today than I was yesterday / than sad."
The acceptability of angrier (synthetic) vs more angry (analytic)
follows from what is deleted and what survives. When the comparison is
across adjectives (angry vs sad), the comparative must be more angry
(analytic) because more is a sentence modifier, not within the AP.
Simple comparative formation (producing angrier) cannot apply
because more and sad are not within the same AP.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Puzzle (D): "*Mary is more than five feet short" vs "Mary is shorter than five feet."
@cite{kennedy-1999} explains this via positive/negative extent boundedness.
Bresnan's syntactic explanation: short rejects definite measure phrase
modifiers (*How short is he? — five feet short), so the QP identity
condition in the clause cannot be satisfied.
Bridge: this connects to measurePhrase_positive_only in Kennedy1999.lean
and admitsMeasurePhrase in Differential.lean.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The syntactic category of the constituent deleted from the than-clause. Bresnan's key insight: the deleted element must be "nondistinct" (same syntactic category) from the head.
This is the 1973 precursor to modern identity conditions in ellipsis (e-GIVENness in @cite{merchant-2001}; SIC in Anand, Hardt & McCloskey 2021).
- qp : DeletionTarget
- ap : DeletionTarget
- np : DeletionTarget
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
The identity condition: deletion succeeds only when the clause constituent and the head have the same syntactic category.
"Nondistinctness" in Bresnan's terms — the deleted constituent must be featurally nondistinct from the head.
Equations
- Phenomena.Comparison.Studies.Bresnan1973.identityHolds head clause = (head == clause)
Instances For
Subdeletion: "The table is longer than the door is wide."
Head = AP (-er much long), deleted = AP (x much wide). Both are APs, so identity holds. The dimensions (length vs width) need not match — only the syntactic category.
This connects to subcomparativeExamples in Kennedy1999.lean and
subcomparative in Intervals.lean.
*John is more than Bill tall fails: head = QP (-er much), but
the matching constituent in the clause is an NP (Bill = that much).
NP ≠ QP, so the identity condition fails.
Bresnan's strongest syntactic claim: ALL comparatives are underlyingly clausal. What appears as a "phrasal" comparative ("taller than Bill") is derived from a full clause by maximal deletion.
This contrasts with the modern consensus (ThanClauseType distinguishes
phrasal from clausal as genuinely distinct). Both analyses yield the
same truth conditions for simple comparatives (proved in
ThanClause.phrasal_clausal_equivalence), so the disagreement is purely
syntactic.
See @cite{bhatt-pancheva-2004} for the modern "genuinely phrasal" view.
- partialDeletion : BresnanThanClauseAnalysis
Full clause with partial deletion: "than Bill is [x much tall]" → "than Bill is" (deletion of AP)
- maximalDeletion : BresnanThanClauseAnalysis
Full clause with maximal deletion: "than Bill is [x much tall]" → "than Bill" (deletion of AP + copula stranding)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Under Bresnan's analysis, the modern phrasal type is just
maximalDeletion of an underlying clause.
Equations
- Phenomena.Comparison.Studies.Bresnan1973.bresnanAnalysisOf Semantics.Degree.ThanClause.ThanClauseType.clausal = Phenomena.Comparison.Studies.Bresnan1973.BresnanThanClauseAnalysis.partialDeletion
- Phenomena.Comparison.Studies.Bresnan1973.bresnanAnalysisOf Semantics.Degree.ThanClause.ThanClauseType.phrasal = Phenomena.Comparison.Studies.Bresnan1973.BresnanThanClauseAnalysis.maximalDeletion
Instances For
Whether an adjective admits definite measure phrase modification ("five feet tall" ✓ vs "*five feet short" ✗).
Bresnan observes this as a syntactic fact; @cite{kennedy-1999} derives it from extent boundedness (positive extents are bounded, negative extents are not). Both predict the same distribution.
Equations
Instances For
Positive adjectives admit measure phrases.
Negative adjectives reject measure phrases.
Bridge to Kennedy 1999: Bresnan's syntactic observation and Kennedy's semantic explanation make the same prediction for simple cases.
Bresnan: short syntactically rejects definite measure QPs, so the
identity condition in comparative deletion cannot be satisfied.
Kennedy: negative extents are unbounded, so equating them with bounded measure phrases is semantically undefined.
The predictions diverge for less short than five feet (Bresnan predicts
OK via little-based QP; Kennedy predicts OK via comparative semantics).
- adjective : String
- polarity : Core.Scale.ScalePolarity
- measurePhraseOk : Bool
- comparativeMeasureOk : Bool
"more than N units Adj" acceptable?
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
Among Bresnan's dimensional adjectives, measure phrase acceptability correlates with positive polarity.
This is the formal content behind puzzle (D) and connects to
measurePhrase_positive_only in Kennedy1999.lean.
Bresnan's paradigms (4) and (5): the Det items form a closed class
that combine with both much and little (mass) / many and few
(count).
Paradigm (4): as/too/that/so/-er + much/little + mass noun
Paradigm (5): as/too/that/so/-er + many/few + count noun
The last row in each paradigm undergoes suppletion:
-er much → more, -er little → less,
-er many → more, -er few → fewer.
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
Verify: every suppletive entry in the paradigm matches the suppletion function.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Verify: every suppletive entry in the count paradigm matches the suppletion function.
enough requires a null Det — *so enough, *too enough, *as enough,
*enougher are all impossible (Bresnan p. 286).
This is formalized as a well-formedness predicate on QPs.
Equations
- qp.isWellFormed = match qp.q with | Phenomena.Comparison.Studies.Bresnan1973.Q.enough => qp.det == Phenomena.Comparison.Studies.Bresnan1973.Det.null | x => true
Instances For
Bresnan's "archicategory" X̄ unifies QP and AP: both have left-recursive specifier + head structure.
QP → (QP) QP, QP → (Det) Q AP → (AP) AP, AP → (Adv) A
This is a direct precursor to X-bar theory (Jackendoff 1977) and ultimately to the Minimalist bare phrase structure. The modern counterpart is the shared structure of DegP and AdjP as functional and lexical projections in the extended AP.
Structural rule from Bresnan (145): X̄ → (X̄) X̄ (recursive expansion) X̄ → (Spec, X̄) X̄ (specifier + head)
- qp : BarCategory
- ap : BarCategory
- np : BarCategory
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Bresnan's analysis is primarily syntactic, but the truth conditions it derives are compatible with the modern degree-semantic consensus.
For a simple adjectival comparative "A is taller than B":
- Bresnan: head = AP [-er much tall], clause deletes matching AP [x much tall]
- Kennedy: μ(A) > μ(B) where μ = height
- Heim: max({d | height(A) ≥ d}) > max({d | height(B) ≥ d})
All three yield the same truth conditions for simple cases.