Rákosi (2019): Reciprocal anaphors in singular constructions in Hungarian #
@cite{rakosi-2019}
Workshop on Cross-Linguistic Semantics of Reciprocals, Utrecht University, 7–8 October 2019. Proceedings edited by Palmieri, Winter & Zwarts (2020).
Core Empirical Generalization #
Hungarian reciprocals (egymás) tolerate morphosyntactically singular antecedents in four construction types (§§3–6), while reflexives (maga/maguk) require morphosyntactic plurality (plural noun head + plural verb agreement + plural anaphor form).
| § | Construction | Syn# | Sem# | Refl(PL) | Recip |
|---|---|---|---|---|---|
| 3 | Quantified NP | SG | PL | ✗ | ✓ |
| 4 | Singular coordinate DP | SG | PL | ✗ | ✓ |
| 5 | Collective noun | SG | PL | ✗ | ✓ |
| 6 | Bound variable (pro) | SG | PL | — | ✓ |
| — | Plural NP (baseline) | PL | PL | ✓ | ✓ |
Theoretical Claim #
The plurality requirement on reciprocal antecedents is semantic, not
morphosyntactic. This follows from the formal semantics of the
anaphoric relations: reciprocity (R) requires per-situation distinctness
(u_ant s ≠ u_pro s), presupposing multiple individuals in the
denotation. Reflexive binding (=) operates via φ-feature agreement,
a morphosyntactic mechanism.
Connections #
Theories/Semantics/Reference/PluralityLicensing.lean— thePluralityRequirementsubstrate (anchored on this paper).Theories/Semantics/Dynamic/PPCDRT/Anaphora.lean— the formal-semantic reciprocity / binding conditions over plural assignments.Fragments/Hungarian/Reciprocals.lean—AntecedentConfig,reciprocalLicensed,pluralReflexiveLicensed, verification.Phenomena/Anaphora/Studies/DalrympleHaug2024.lean§2 — the bound-variable case (§6 here) is also discussed there as evidence for the relational analysis of reciprocal scope.Phenomena/Anaphora/Coreference.lean— thereciprocalPatternnotes that syntactically singular antecedents are possible.
The paper's core generalization: reciprocals need semantic plurality, reflexives need morphosyntactic plurality.
This is not stipulated — it is derived from anaphorPluralityReq
in the theory layer, which in turn follows from the formal semantics
of binding (=) vs. reciprocity (R).
All four singular constructions license reciprocals.
No singular construction licenses the plural reflexive.
§2 rules out a potential confound: Hungarian "inclusive reference" reflexives (1SG subject + "ourselves") look like singular antecedent
- plural reflexive, but these are NOT bound variables.
Evidence: under csak ('only'), the inclusive reflexive gets a referential reading ("for us"), not a bound-variable reading ("for themselves"). True anaphors (matching-φ reflexives and reciprocals) DO get bound readings under csak.
The reciprocal is NEVER licensed in inclusive reference: "*Sokszor sajnálom egymás-t" is ungrammatical — the 1SG antecedent is not semantically plural.
- boundUnderOnly : Bool
Can the inclusive reflexive be bound under "only"?
- reciprocalPossible : Bool
Can the reciprocal appear in the inclusive construction?
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Rakosi2019.instReprInclusiveRefData = { reprPrec := Rakosi2019.instReprInclusiveRefData.repr }
Equations
- Rakosi2019.inclusiveReflexive = { boundUnderOnly := false, reciprocalPossible := false }
Instances For
Inclusive reflexives are not true anaphors: they don't bind under only, and the reciprocal is categorically excluded.
§3: Quantified antecedents. Hungarian quantified NPs are morphologically singular and take 3SG verbs.
(8a) A két gyerek jól érezte magá-t/*maguk-at. 'The two children felt well.' (SG reflexive only)
(9a) A szobában három kisgyerek kergeti egymás-t. 'Three little children are chasing each other.' (reciprocal OK)
§4: Singular coordinate DPs.
(11a) Kati és Éva kihúzta magát/*magukat. 'Kati and Éva drew themselves up.' (3SG verb → SG reflexive only)
(12) Kati és Éva látta/látták egymás-t a tükörben. 'Kati and Éva saw each other.' (reciprocal OK with SG or PL verb)
§5: Collective noun antecedents.
(14) A személyzet fáradt volt/*voltak. 'The staff was tired.' (collective nouns: 3SG agreement only)
(15a) A személyzet riadtan nézte egymás-t. 'The staff were watching each other frightened.' (reciprocal OK)
(16) Az egész család jól érezte magá-t/*maguk-at. 'The whole family enjoyed themselves.' (SG reflexive only)
§6: Bound variable antecedent.
(17) Péter és Éva az-t gondolja, hogy (*ő) szereti egymás-t. 'Péter and Éva think that they love each other.' (pro-dropped 3SG embedded subject, reciprocal OK, wide scope only)
This case is also discussed in @cite{dalrymple-haug-2024} §2 as evidence for the relational analysis: the singular bound pronoun forces binding (=), yielding the I-reading.
The semantic justification: reciprocity requires distinct individuals, which is a denotation-level property. A semantically-plural but syntactically-singular antecedent provides the needed distinct individuals. A truly singular (atomic) antecedent does not — which is why "*A gyerek kergeti egymás-t" (the child chases each other) is ungrammatical even though the verb is 3SG.
The formal semantics connection: reciprocity (R) presupposes
multiple individuals in the range of the discourse referent function.
This is derived in PluralityLicensing.lean:
reciprocity_implies_multiple_individuals. The PPCDRT version uses
plural information states; the witness here is the contrapositive —
if both anaphor and antecedent are mapped to the same value 0 in a
singleton state, the distinctness clause of reciprocityCond rules
out reciprocity.
Binding (=) is compatible with a singleton state where both drefs point to the same value — explaining why reflexives don't impose a semantic plurality requirement.
egymás is morphologically invariable — it bears no number feature. This is consistent with the claim that its plurality requirement is semantic, not morphosyntactic: it doesn't participate in φ-agreement.
The reflexive DOES participate in φ-agreement: maga (SG) vs. maguk (PL). The anaphor's number must match the verb's agreement, confirming that reflexive licensing is morphosyntactic.
The morphological invariance of egymás predicts it should be insensitive to verb agreement number — and it is: reciprocals are grammatical with both SG and PL verbs when the antecedent is a coordinate DP (ex. 12: "Kati és Éva látta/látták egymás-t").