Haug and Dalrymple (2020): Reciprocity: Anaphora, scope, and quantification #
Haug and Dalrymple give a relational analysis of reciprocity in
Partial Plural Compositional DRT. The reciprocal contributes group identity with its
antecedent and pointwise distinctness ((41), PPCDRT.reciprocityCond); the scope
ambiguity of a reciprocal in a complement clause is the ambiguity of any plural anaphor
between group identity (PPCDRT.groupIdentityCond) and binding (PPCDRT.bindingCond)
with the matrix subject, read through the distribution of §2.3 under which the antecedent
escapes the distribution operator. The sample output states of the paper are transcribed
as plural information states (ofRows) and the conditions checked on them: the two
readings of the lawyers' secretary (24), (31), the narrow and wide readings of the girls'
belief (49), (51), (53), (55), the crossed reading (56), the modal (69), the mixed
construal of the Cheyenne affix (78), multiple reciprocals (85), (86), the forks (93), the
sailors (96), and the wide construal with three girls (136).
no_low_reciprocal_under_binding derives the empty cell of §3.3: a bound antecedent
leaves no plurality for a reciprocal inside the distribution.
The paper's relation to cumulation is stated in general. Pointwise conditions on plural
states give cumulative readings (§2.1, cumulative_iff_exists_relCond), with the truth
conditions of Sternefeld's cumulation operator (§4.1, cumulation_iff_exists_relCond).
A simple reciprocal sentence is true exactly when its antecedent is weakly reciprocal under the
verb (§2.4, weakReciprocity_iff_exists_reciprocalDRS), which is Sternefeld's reading (76)
(cumulation_iff_exists_reciprocalDRS). Without distribution, and whatever the verb, the
dependency between the antecedent and the reciprocal is weakly reciprocal
(weakReciprocity_dep_of_reciprocityCond), of which the forks are an instance (forks_weak).
Quantified antecedents (§5) get the two readings of (99), RefSetReading over a maximal
reciprocal subset and MaxSetReading over the participants, and the supervaluation truth
value of (109), truthValue, with maxSetReading_of_refSetReading for upward monotone
determiners and the street and club scenarios of §5.1–§5.2. Maximize Anaphora (128) is
MaximizesAnaphora; it selects the states whose dependency contains every pair the verb
allows (maximizesAnaphora_iff), hence Strong Reciprocity when the verb allows every pair
(strongReciprocity_dep_iff), the strong reading (126b) over the minimal state (126a)
(maximize_strong), and it maximizes multiple reciprocals pairwise (§6.2).
Implementation notes #
- Distribution is represented by the set
Δof aPPCDRT.PPDRSCond; the operatorsδ,Tandthinkof (14), (46) are not defined, and each sample state is checked under theΔits DRS induces. Themaxoperator of (97) is likewise replaced by the static readings it yields. - A DRS is true when some output state satisfies its conditions.
relCondasks every state to value its discourse referents, which the partial introduction (6), (20) guarantees: the statesPPCDRT.introPartialreaches from the empty state{⊥}are exactly those whose rows value the introduced drefs (introPartial_comp_iff), and truth reached from{⊥}agrees with truth at some state (weakReciprocity_iff_exists_introPartial). ReciprocalDRSfixes the antecedent's plurality,∪u' = X, as a definite antecedent does; (125b) has the pointwise restrictorboy(u₁)instead.boysKnowtakesXto be the three boys.- The pairs
R_uof (127) are the dependencyPCDRT.dep, taken antecedent first so that they line up with the verb's arguments; the order does not affect maximization. - Worlds are values of the same domain as individuals, at the discourse referent
w. - The antecedence condition for the second reciprocal of (85b) is read with
u₂, as the indices of (85a) and the state (85c) give.
TODO #
- The distribution and maximization relations
δ_uandmax^uof (14) and (97), so that the sample states are derived from the DRSs rather than transcribed; introduction, (6) and (20), isPPCDRT.introPartial. - With the pointwise restrictor
∪u' ⊆ Aof (125b) in place of∪u' = X, Maximize Anaphora selects the states whose dependency is full on a maximal weakly reciprocal subset ofA, the shape of the reference sets of §5 (IsRefSet).
References #
- D. T. T. Haug and M. Dalrymple, Reciprocity: Anaphora, scope, and quantification (2020)
- D. T. Langendoen, The logic of reciprocity (1978)
- W. Sternefeld, Reciprocity and cumulative predication (1998)
- S. Beck and U. Sauerland, Cumulation is needed: A reply to Winter (2000) (2000)
- J. Dotlačil, Reciprocals distribute over information states (2013)
- S. E. Murray, Reflexivity and reciprocity with(out) underspecification (2008)
- M. Križ, Aspects of homogeneity in the semantics of natural language (2015)
- L. Champollion, D. Bumford and R. Henderson, Donkeys under discussion (2019)
Sample states #
The individuals and worlds of the sample states.
- girl1 : Ind
- girl2 : Ind
- tracy : Ind
- chris : Ind
- matty : Ind
- world1 : Ind
- world2 : Ind
- world3 : Ind
- lawyer1 : Ind
- lawyer2 : Ind
- lawyer3 : Ind
- secretary1 : Ind
- secretary2 : Ind
- secretary3 : Ind
- picture1 : Ind
- picture2 : Ind
- picture3 : Ind
- fork1 : Ind
- fork2 : Ind
- fork3 : Ind
- sailor1 : Ind
- sailor2 : Ind
- sailor3 : Ind
- ship1 : Ind
- ship2 : Ind
- ship3 : Ind
- child1 : Ind
- child2 : Ind
- child3 : Ind
- boy1 : Ind
- boy2 : Ind
- boy3 : Ind
Instances For
Equations
- HaugDalrymple2020.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- HaugDalrymple2020.instFintypeInd = { elems := { val := ↑HaugDalrymple2020.Ind.enumList, nodup := HaugDalrymple2020.Ind.enumList_nodup }, complete := HaugDalrymple2020.instFintypeInd._proof_1 }
Equations
- HaugDalrymple2020.instReprInd = { reprPrec := HaugDalrymple2020.instReprInd.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
A row of a sample state, from its defined discourse referents.
Equations
- HaugDalrymple2020.row l u = (List.find? (fun (x : ℕ × E) => x.1 == u) l).elim ⊥ fun (p : ℕ × E) => ↑p.2
Instances For
The plural information state with the given rows.
Equations
- HaugDalrymple2020.ofRows rows = {g : PartialAssign ℕ HaugDalrymple2020.Ind | g ∈ rows}
Instances For
The values of u summed over the Δ-class of s, as a finset.
Equations
- HaugDalrymple2020.sumRows rows Δl s u = {d : HaugDalrymple2020.Ind | ∃ t ∈ rows, (∀ v ∈ Δl, t v = s v) ∧ t u = ↑d}
Instances For
The collective condition n-atoms(∪u) of (9) under distribution holds when, in every
state, the values of u summed over the state's class number n ((27), (28b)).
Equations
- HaugDalrymple2020.atomsCond n u S Δ = ∀ s ∈ S, (PCDRT.value u (PPCDRT.eqClass S Δ s)).ncard = n
Instances For
A lexical relation as a condition ((7), (27)) holds when the state is nonempty and every
state gives u and v values that P relates. Distribution does not reach individual discourse
referents, so Δ is idle.
Equations
- HaugDalrymple2020.relCond P u v S x✝ = (Set.Nonempty S ∧ ∀ s ∈ S, ∃ (a : E) (b : E), s u = ↑a ∧ s v = ↑b ∧ P a b)
Instances For
Group identity read inside the distribution operator of (14), summing both sides over the class. This is the reading under which (23b) is not representable (§2.3).
Equations
- HaugDalrymple2020.distributedGroupIdentityCond uAnaph uAnt S Δ = ∀ s ∈ S, PCDRT.value uAnaph (PPCDRT.eqClass S Δ s) = PCDRT.value uAnt (PPCDRT.eqClass S Δ s)
Instances For
Anaphora under distribution (§2.3) #
The lawyers hired secretaries all of them liked, ∪u₃ → ∪u₁ under δ_{u₁} (31c).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Read as (24), (23a) has they bound by the lawyers, one secretary per lawyer under
δ_{u₁}, and no group identity under the distribution.
Read as (31), (23b) has ∪u₃ → ∪u₁ escape δ_{u₁}, with one secretary per lawyer and
no binding. The same state fails group identity read inside the distribution operator of
(14), which is why (23b) needs the distribution of §2.3.
Cumulative readings and Weak Reciprocity (§1, §2.1, §2.4, §4.1) #
The state with a row for each pair of D, valuing u and v as its two components, so
that its dependency between u and v is D.
Equations
- HaugDalrymple2020.ofDep u v D = (fun (p : E × E) => HaugDalrymple2020.row [(u, p.1), (v, p.2)]) '' D
Instances For
Under P(u, v) the dependency between u and v lies in P, with the values of u as
its domain and the values of v as its codomain.
By §2.1, (12), cumulative readings are the default. A state satisfying P(u, v) relates the
values of u to the values of v cumulatively in the sense of [BS00], and any
two pluralities so related are the values of such a state.
By §4.1, (75), the cumulation operator ** that [Ste98] applies to the predicate and
the pointwise conditions of Plural CDRT give the same truth conditions.
In the DRS of a reciprocal sentence whose antecedent denotes X, as in (40b) and (125b),
the antecedent u' sums to X, P relates u' to the reciprocal u in every state, and u
is reciprocal to u' ((42)).
Equations
- HaugDalrymple2020.ReciprocalDRS P X u u' S = (PCDRT.value u' S = X ∧ HaugDalrymple2020.relCond P u' u S ∅ ∧ PPCDRT.reciprocityCond u u' S ∅)
Instances For
The reciprocal DRS is the pointwise condition of P with non-identity conjoined, with X
in both argument positions. Group identity puts X in the second, and the distinctness
presupposition contributes the non-identity.
As §2.4 puts it, "This proposal, like Dotlačil's, makes Weak Reciprocity the basic reading."
Some state verifies the DRS of the Xs P each other exactly when X is weakly reciprocal
under P ([Lan78]). The reciprocal is cumulative predication of P with non-identity
conjoined, which is Langendoen's comparison of the women pointed at each other with the women
released the prisoners (§1).
By §4.1, (76), [Ste98]'s weak reciprocal, ** of P with non-identity conjoined
served the same plurality twice, and the PPCDRT reciprocal (42) agree on simple reciprocal
sentences; they differ in architecture, not in these truth conditions.
Introducing u' and then u at the empty state {⊥}, (6) and (20), reaches exactly the
nonempty states whose rows value u' and u and nothing else.
Entered by introducing u' and u at {⊥} ((6), (20)), the DRS of the Xs P each other has
an output verifying it exactly when X is weakly reciprocal under P. A verifying state
restricted to its u' and u columns is reachable and still verifies the DRS.
By §2.4, §4.5, whatever the verb, the dependency between the antecedent and a reciprocal
without distribution is weakly reciprocal on the antecedent's plurality. The forks of (92),
whose verb takes ∪u₂ collectively, are "a special case of weak reciprocity" in this sense.
Reciprocal scope (§3) #
Each girl thought "we will win", ∪u₂ → ∪u₁ under δ_w (49).
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §3.1, the two readings of (44) come apart under the distribution the attitude verb induces, (49) being group identity without binding and (51) binding without group identity.
Each girl thought "we saw each other", the reciprocal inside the belief (53).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Each girl thought "I saw her", the reciprocal and its antecedent lifted to the matrix DRS (55).
Equations
- One or more equations did not get rendered due to their size.
Instances For
In the crossed reading of (56), girl1 thought girl2 saw girl1 and girl2 thought girl1 saw girl2.
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §3.2, (55) satisfies (54), binding of the pronoun and reciprocity in the matrix.
By §3.3, the crossed state satisfies (56), group identity of the pronoun with reciprocity in the matrix, but not the binding of (54).
By §3.3, a bound antecedent cannot cooccur with a low reciprocal. If u₂ is bound by
u₁, the distribution runs over u₁, and the reciprocal u₃ is reciprocal to u₂ under
that distribution, no state can assign u₂ a value, since the value must be covered by u₃
within the class, where distinctness excludes it.
Example (69) distributed over two accessible worlds, with the reciprocal lifted above ◇.
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §3.4, with the reciprocal above the modal, every accessible world contains both
directions of beat, the contradiction that makes (64) strange.
Underspecification, multiple reciprocals, subgroups, collectives (§4) #
The mixed construal of the Cheyenne affix (78c) has one child scratching herself and two scratching each other.
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §4.2, the mixed construal satisfies the underspecified meaning (79b) but neither reciprocity nor reflexive binding.
Each girl gave the other a picture of herself, the second reciprocal anteceded by the first (85c).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Each girl gave the other a picture of the other, both reciprocals anteceded by the subject (86c).
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §4.4, the two readings of (84) are the two antecedents of the second reciprocal; each state satisfies its own DRS and fails the other's distinctness condition.
The forks propped against each other, each on one other (93).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The three forks.
Equations
Instances For
By §4.5, the forks reading is a special case of weak reciprocity. The minimal state (93) is reciprocal, and its dependency is weakly but not strongly reciprocal on the three forks, since each fork is propped against another, not against every other.
The sailors worked together on each other's ships (96c).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Quantified antecedents (§5) #
A reference set (101) is a subset of the restrictor A over which R is strongly
reciprocal (fn. 24), of the largest cardinality any such subset has, as the operator of (97)
maximizes. Maxima need not be unique (fn. 18).
Equations
- HaugDalrymple2020.IsRefSet R A Y = (Y ∈ A.powerset ∧ Reciprocal.StrongReciprocity R Y ∧ ∀ Z ∈ A.powerset, Reciprocal.StrongReciprocity R Z → Z.card ≤ Y.card)
Instances For
In the reference-set reading, (101) and (110a), the determiner holds of the restrictor and a reference set.
Equations
- HaugDalrymple2020.RefSetReading Q R A = ∃ Y ∈ A.powerset, HaugDalrymple2020.IsRefSet R A Y ∧ Q A Y
Instances For
The participants (105) are the members of A bearing R to some other member.
Equations
- HaugDalrymple2020.participants R A = {a ∈ A | ∃ b ∈ A, b ≠ a ∧ R a b}
Instances For
In the maximal-set reading, (104) and (110b), the reciprocal ranges over the whole restrictor and the determiner holds of the restrictor and the participants.
Equations
- HaugDalrymple2020.MaxSetReading Q R A = Q A (HaugDalrymple2020.participants R A)
Instances For
The two ranges of the reciprocal in (99).
Instances For
Equations
- HaugDalrymple2020.instDecidableEqRange x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- HaugDalrymple2020.instReprRange = { reprPrec := HaugDalrymple2020.instReprRange.repr }
The reading of a quantified reciprocal sentence at a range.
Equations
Instances For
Each member of a reference set participates.
By §5.2, for a determiner upward monotone in its scope, the reciprocal relation holds over the maximal set if it holds over a reference set, so the reference-set reading determines truth and the maximal-set reading falsity.
Equations
- One or more equations did not get rendered due to their size.
A quantified reciprocal sentence is true iff true at both ranges, false iff false at both, and neither otherwise, the supervaluation over the two precisifications (109).
Equations
Instances For
The five inhabitants of the street of (100) and (102).
Equations
- HaugDalrymple2020.Person = Fin 5
Instances For
Proportional most holds when the scope has more than half of the restrictor.
Equations
- HaugDalrymple2020.most A B = (A.card < 2 * B.card)
Instances For
Proportional few holds when the scope has less than half of the restrictor.
Equations
- HaugDalrymple2020.few A B = (2 * B.card < A.card)
Instances For
The first three inhabitants know each other and the other two know nobody (102).
Equations
- HaugDalrymple2020.clique a b = (a ≠ b ∧ ↑a < 3 ∧ ↑b < 3)
Instances For
The intermediate scenario of §5.2, in which two pairs know each other and one person knows nobody.
Equations
- HaugDalrymple2020.pairs a b = (a ≠ b ∧ (↑a < 2 ∧ ↑b < 2 ∨ 2 ≤ ↑a ∧ ↑a < 4 ∧ 2 ≤ ↑b ∧ ↑b < 4))
Instances For
Equations
- HaugDalrymple2020.instDecidableRelPersonClique x✝¹ x✝ = { decide := decide (x✝¹ ≠ x✝) && decide (↑x✝¹ < 3 ∧ ↑x✝ < 3), reflects_decide := ⋯ }
Equations
- HaugDalrymple2020.instDecidableRelPersonPairs x✝¹ x✝ = { decide := decide (x✝¹ ≠ x✝) && decide (↑x✝¹ < 2 ∧ ↑x✝ < 2 ∨ 2 ≤ ↑x✝¹ ∧ ↑x✝¹ < 4 ∧ 2 ≤ ↑x✝ ∧ ↑x✝ < 4), reflects_decide := ⋯ }
Equations
- HaugDalrymple2020.instDecidableRelFinsetPersonMost x✝¹ x✝ = { decide := x✝¹.card.succ.ble (2 * x✝.card), reflects_decide := ⋯ }
Equations
- HaugDalrymple2020.instDecidableRelFinsetPersonFew x✝¹ x✝ = { decide := (2 * x✝.card).succ.ble x✝¹.card, reflects_decide := ⋯ }
By (110) and (111), "most people know each other" is true and "few know each other" false in the clique scenario of (102), and both are neither in the scenario of two pairs, which Kamp and Reyle judged arguably true.
Maximize Anaphora (§6) #
A state maximizes anaphora (128) when, among the states a DRS K admits, its set R_u of
pairs (127), the dependency between antecedent and anaphor, is not properly included in another
admitted state's.
Equations
- HaugDalrymple2020.MaximizesAnaphora K uAnaph uAnt S = (K S ∧ ∀ (S' : PluralAssign ℕ E), K S' → ¬PCDRT.dep uAnt uAnaph S ⊂ PCDRT.dep uAnt uAnaph S')
Instances For
A reciprocal DRS realizes only pairs of distinct members of X that P relates.
Maximize Anaphora (§6, (128)) selects exactly the states of a reciprocal DRS whose dependency
is every pair of distinct members of X that P relates.
By §6, full maximization yields Strong Reciprocity. Under Maximize Anaphora the dependency
makes X strongly reciprocal exactly when P relates any two distinct members of X.
The three boys of (125).
Equations
Instances For
A minimal state for (125), (126a).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The strong reciprocal reading (126b).
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §6, the minimal state (126a) verifies (125), but Maximize Anaphora rejects it and selects (126b), whose dependency is the strong reciprocal reading.
By §6.2, the classmates gave each other pictures of each other, each maximizing pairwise, so that each gave a picture of herself to each of the others.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The three classmates.
Equations
Instances For
By §6.2, both reciprocal relations of (134) are maximal, strongly reciprocal on the classmates, without any all-triples row, in which a classmate gives a picture of a third classmate.
In the wide construal of (135), each of Tracy, Matty and Chris believes that she praised the two others (136).
Equations
- One or more equations did not get rendered due to their size.
Instances For
By §6.3, (136) is the wide reading, binding of the pronoun with reciprocity in the matrix, and within each girl's world the reciprocal covers exactly the two others.