Dalrymple & Haug (2024): Constraints on Reciprocal Scope #
@cite{dalrymple-haug-2024}
Linguistic Inquiry, Early Access. DOI: 10.1162/ling_a_00546.
Properties of the local antecedent of the reciprocal (the embedded-clause pronoun coreferent with the matrix subject) determine reciprocal scope. This paper systematically surveys five construction types and shows that the relational analysis of reciprocals makes correct predictions in all cases, while the quantificational analysis fails for distributive operators (§5) and logophoric antecedents (§6).
Construction Types Surveyed #
| § | Construction | Narrow | Wide | Both agree? |
|---|---|---|---|---|
| 2 | Bound antecedent (Hungarian) | ✗ | ✓ | ✓ |
| 2 | Nonbound antecedent (Japanese) | ✓ | ✗ | ✓ |
| 3 | Collective conjunct | ✓ | ✗ | ✓ |
| 4 | Partial control | ✓ | ? | ✓ |
| 4 | Exhaustive control, collective | ✓ | ✗ | ✓ |
| 4 | Exhaustive control, non-coll. | ✗ | ✓ | ✓ |
| 5 | Distributive operator | ✓ | ✓ | ✗ (quant. ✗) |
| 6 | Logophoric antecedent (Wan) | ✓ | ✗ | ✗ (quant. ✗) |
Connections #
Theories/Semantics/Reference/Reciprocals.lean— the three anaphoric relations (=, ∪, R) and prediction functionsTheories/Semantics/Lexical/Plural/Distributivity.lean— distributive operators (§5 shows each does NOT block wide scope, contra @cite{heim-lasnik-may-1991})Fragments/Hungarian/Reciprocals.lean— Hungarian egymás with singular null pronoun antecedent (§2, @cite{rakosi-2019})Fragments/Wan/Reciprocals.lean— Wan logophoric reciprocal data (§6)Core/Discourse/Logophoricity.lean— Sells (1987) logophoric rolesPhenomena/Control/Studies/Landau2015.lean— control tier distinction (exhaustive vs. partial) relevant to §4
A reciprocal scope judgment: which readings are available for a particular construction type.
- construction : String
Description of the construction
- example_ : String
Example sentence
- available : List Semantics.Reference.Reciprocals.RecipScope
Available scope readings
- section_ : ℕ
Which section of the paper
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Hungarian: singular null pronoun antecedent forces bound reading (=), yielding only wide scope.
(10) Péter és Éva az-t gondolja, hogy szereti egymás-t. 'Péter and Éva think that [they] love each other.' → Wide scope only (I-reading)
Equations
- One or more equations did not get rendered due to their size.
Instances For
Japanese: zibun-tati (plural reflexive) resists bound reading, favoring group identity (∪). As antecedent of reciprocal, only narrow scope available.
(11) John to Mary ga [zibun-tati ga otagai o mi-ta to] omow-ta. 'John and Mary thought that selves saw each other.' → Narrow scope only (we-reading)
Equations
- One or more equations did not get rendered due to their size.
Instances For
When the reciprocal VP is coordinated with a collective predicate, only narrow scope is available. Wide scope gives the local antecedent an individual denotation, which cannot satisfy the collectivity requirement.
(12) The girls hoped that they would [meet at the tennis court] and [defeat each other]. → Narrow scope only
Equations
- One or more equations did not get rendered due to their size.
Instances For
The canonical wide-scope-only case in control constructions (@cite{higginbotham-1980}).
(13) They wanted to visit each other. → Wide scope only (I-reading)
This has been "generally accepted" since @cite{higginbotham-1980}. Note that want is actually partial control per @cite{landau-2015}, so the scope fixing here may be due to pragmatic factors rather than syntactic constraints on control type.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Exhaustive control with a collectively interpreted controller rules out wide scope (collective interpretation requires plurality).
(14) They decided to keep each other's comments confidential. → Narrow scope available (collective "decided")
@cite{heim-lasnik-may-1991} claim this has two readings (collective narrow, distributive wide), but the distinction is hard to verify for many verbs.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Partial control verbs (want, hope) with a singular controller allow narrow scope because PRO can denote a superset of the controller.
(15a) I asked a girl who I liked if she wanted to get to know each other better. (15b) I vow to keep reminding you McDonald's is unhealthy ... because I want to live long, happy lives by each other's side. → Narrow scope available
These are attested corpus examples. Substituting exhaustive control verbs (try, manage) makes the sentences worse (16a-b), confirming the role of partial control. The matrix argument is singular, so a collective interpretation is impossible — narrow scope arises purely from the partial-control PRO denoting a superset.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Exhaustive control with non-collective interpretation: wide scope only.
(17) Unbeknownst to each other, Tracy and Chris intended to help each other. → Wide scope only
"Unbeknownst to each other" forces a distributive reading of the matrix subject (each individual is unaware), so the controller is not interpreted collectively.
Equations
- One or more equations did not get rendered due to their size.
Instances For
@cite{heim-lasnik-may-1991} claim that (18a) is unambiguous (narrow only) and that (18b) is ungrammatical:
(18a) They each think they are taller than each other. (18b) *They each examined each other.
Their reasoning: on the quantificational analysis, distributive each cannot apply to the already-distributed NP each inside each other. This is "a commonplace in the literature on distributivity" (@cite{champollion-2016}).
However, corpus examples with each of them / each and a reciprocal antecedent are plentiful (19-20, 23-26), and both readings are available (24-25 for narrow, 25-26 for wide).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Contra @cite{heim-lasnik-may-1991}, a distributive operator (each) in the matrix clause does NOT block wide scope. Corpus data shows both readings are available.
(24) They each liked each other. [narrow: mutual knowledge] (25) They each liked each other before. [wide: I-reading] (27b) They each think they liked each other. [narrow] (27c) They each think they liked each other. [wide]
The quantificational analysis incorrectly predicts that applying a distributor to an already-distributed NP should be impossible. The relational analysis correctly allows both: each other is a pronoun, not a quantified NP, so there is no double-distribution problem. Distributive each can access the group denoted by the antecedent even if we distribute on that antecedent (@cite{haug-dalrymple-2020} §2.3).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The each...different diagnostic (exx. 21-23) shows that each in they each [V] each other is not vacuous: it licenses internal readings of different that bare each other cannot.
(21a) The men told each girl a different story. → Internal reading available (different story per girl) (21b) The men told each other a different story. → External reading only (not: different per pair member) (22) The men each told each other a different story. → Internal reading emerges with each present
This proves that higher each has genuine semantic content (it distributes on the antecedent), contra the HLM prediction that it should be impossible.
Equations
- One or more equations did not get rendered due to their size.
Instances For
In Wan (Mande), when the antecedent of the reciprocal is a logophor, only narrow scope is available.
(28) wì mù tēŋ gé mɔ̄ á ē ɔ̄ŋ lɔ̄ lé 'All the animals say they-LOG will eat each other.' → Narrow scope only (logophor + reciprocal)
The wide scope reading IS available with an ordinary (non-logophoric) pronoun (32), confirming that logophoricity is the constraining factor.
The relational analysis predicts this: on the wide scope reading, the embedded subject must be interpreted in the matrix clause for accessibility, but a logophor is confined to the report context. The quantificational analysis predicts both readings should be available since the quantifier scopes independently of logophoricity.
Equations
- One or more equations did not get rendered due to their size.
Instances For
With an ordinary (non-logophoric) pronoun, the wide scope reading IS available in Wan.
(32) wì mù tēŋ tú gé à ɔ̄ŋ lɔ̄ lé 'They all say they-3PL are going to eat each other.' → Both readings available
Equations
- One or more equations did not get rendered due to their size.
Instances For
All scope judgments from the paper.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Antecedent properties for each construction.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Japanese zibun-tati resists bound readings (@cite{nishigauchi-1992}), forcing group identity (∪) and thus narrow scope only.
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
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
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
Hungarian bound antecedent: relational predicts wide only.
Japanese zibun-tati forces group identity (∪), yielding narrow scope only. Both analyses agree on this.
Collective conjunct: relational predicts narrow only.
Exhaustive control, collective: relational predicts narrow only.
Exhaustive control, non-collective: relational predicts wide only.
Distributive operator: relational predicts both readings. The distributor is orthogonal because each other is a pronoun.
Logophoric antecedent: relational predicts narrow only.
Both analyses agree on Japanese zibun-tati: narrow only.
The quantificational analysis makes the SAME prediction as the relational analysis for bound antecedents.
The quantificational analysis makes the SAME prediction for collective conjuncts.
The quantificational analysis agrees on exhaustive control.
DIVERGENCE 1 (§5): The quantificational analysis incorrectly predicts that distributive each blocks wide scope for reciprocals. It predicts narrow scope only, but empirically both readings are attested (exx. 19-20, 24-26).
The relational analysis correctly predicts both readings are available, because each other is a pronoun, not a quantified NP.
DIVERGENCE 2 (§6): The quantificational analysis fails to restrict scope for logophoric antecedents. It predicts both readings should be available, but empirically only narrow is attested.
On the quantificational analysis, the quantifier part of each other scopes independently of whether the embedded subject is logophoric. The relational analysis correctly predicts narrow only, because the logophor is confined to the report context and the reciprocal's R-relation cannot "drag" its antecedent out.
The relational analysis matches the empirical data in ALL cases where the two analyses diverge. The quantificational analysis is empirically inadequate for distributive operators and logophoric antecedents.
Hungarian egymás is formally distinct from the reflexive maga, per the fragment data. This distinction matters: reflexives and reciprocals have different scope possibilities.
Hungarian allows singular antecedents for the reciprocal, which forces wide scope (the singular null pronoun must be bound).
Wan logophoric pronoun mɔ̄ is formally distinct from the ordinary 3pl pronoun tú. The scope constraint is specific to logophoric antecedents — wide scope IS available with the ordinary pronoun.
Wan reflexive ē is distinct from the logophoric pronoun mɔ̄. The reciprocal construction uses REFL + RECIP morphology (ē ɔ̄ŋ), while the logophoric pronoun (mɔ̄) is the subject of the embedded clause.
The Wan logophoric pronoun satisfies at least the pivot role in
@cite{sells-1987}'s hierarchy, connecting this fragment to the
logophoricity theory in Core/Discourse/Logophoricity.lean.
The bound antecedent case (Hungarian) uses the bindingSem relation,
which implies groupIdentitySem. Since binding forces an individual
denotation for the local antecedent, only wide scope is possible.
This connects the scope predictions to the formal semantics
of @cite{haug-dalrymple-2020} §§2.2--3.
The logophoric restriction (Wan) is modeled at the enum level:
isLogophoric = true restricts to narrow scope, which corresponds to
the groupIdentitySem relation. The logophor's discourse referent is
confined to the report context, preventing the bindingSem relation
that would yield wide scope.