Bubnov (2026): Not all coexpressions are syncretisms: Limiting Nanosyntax #
Indefinite pronouns coexpress the specific known, specific unknown and non-specific functions in the patterns AAA, ABB, AAB and ABC and never in the pattern ABA. Dekier derives the gap from a containment hierarchy whose most complex layer is the specific known one. Bubnov objects that a hierarchy whose higher layers had exponents of their own would show morphological containment, one marker built on another, which no indefinite paradigm does, and argues that the semantic account of Degano and Aloni explains the gap better. There a form carries one restriction on the variation and constancy of the indefinite's value and is used wherever the restriction is met, so coexpression is underspecification rather than syncretism, the unattested pattern is the type whose two restrictions exclude each other and can only be disjoined, and the complex type is the specific unknown one, which no single restriction states.
The accounts come apart on diachrony. Semantically a change weakens a restriction, and every attested change does. On a hierarchy the change the theory is known to allow is the loss of a lexical entry, after which a surviving entry also spells out layers the lost one took from it, all of them below the layers it had. A specific unknown form has become epistemic by gaining the non-specific function and a non-specific form has become epistemic by gaining the specific unknown one, so entry loss derives at most one of the two changes, whichever way the hierarchy is ordered.
Main results #
containment_of_two_exponents,no_containment_of_bottom_entries: a layer with an exponent of its own is a suffix on the form below it, and Dekier's lexicons have no such layer.ordConnected_coverage: a hierarchy ordered either way along the map gives each form an interval of the uses, so it excludes ABA as the semantic account does.no_hierarchy_derives_both_changes,entry_loss_is_weakening: on no hierarchy does entry loss derive both attested changes to the epistemic type, and what it derives is a semantic weakening.specificUnknown_not_atomic: no single restriction states the specific unknown type.coexpression_is_underspecification: each form of the four coexpression patterns is a single type.
TODO #
The paper calls the change from a specific to an unmarked form attested in one sentence of its comparison and unattested in the next, in the discussion of entry loss and in a footnote on the reversed hierarchy. It is taken as unattested here.
References #
Morphological containment #
A structure of two layers that is not spelled out by one exponent is spelled out by the exponent of its lower layer and a suffix. Were a higher layer of the hierarchy to have an exponent of its own, the marker of one function would thus be built on the marker of another, the containment no indefinite paradigm shows.
A lexicon like Dekier's, every entry of which stores the lowest layer, gives each layer a single exponent under the lexicalization algorithm, the one its spellout has there, so no marker is built on another.
Complexity #
No single atom states the specific unknown requirement, whatever its variables, since a constancy atom is closed under subteams, a variation atom under superteams, and the requirement under neither. On the semantic account the specific unknown type is thus the complex one, the specific known and the non-specific type each being a single atom, where the hierarchy makes the specific known layer the most complex.
Diachrony as weakening #
The attested changes weaken the restriction, so the form comes to cover more of the map. A specific unknown form becomes epistemic, a non-specific form becomes epistemic, as German irgend- did ([AP15]), and an epistemic form becomes unmarked.
Diachrony as the loss of an entry #
A hierarchy assigns each function a layer, and a lexicon then gives a form the functions at whose layers it spells the form out.
The functions at which a lexicon spells out a form.
Equations
- Bubnov2026.coverage ℓ L φ = {u : Indefinite.SpecificityFunction | Morphology.Containment.spellout L (ℓ u) = some φ}
Instances For
A hierarchy ordering its layers as the map orders the uses, Dekier's or its reverse, complies
with the ban on ABA, since it gives each form an interval of the uses and so only a type whose
requirement is convex (IndefiniteType.ordConnected_requires).
Entry loss derives the change from one type to another on a hierarchy when some lexicon gives a form the first type and, with some of its entries lost, the second.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A form keeps its functions when other entries are lost.
A function gained through entry loss lies below every function the form had.
A change entry loss derives is a semantic weakening, so the hierarchy predicts a part of what the semantic account predicts.
On no hierarchy does entry loss derive both attested changes to the epistemic type. A specific unknown form gains the non-specific function and a non-specific form the specific unknown one, and each gain would put the gained function's layer below the other's. This holds of Dekier's hierarchy and of its reverse alike.
On Dekier's hierarchy the loss of an entry A storing the non-specific layer makes the
specific unknown form B epistemic, (3) and (4). Entry loss also derives two unattested
changes there, a specific known form becoming specific and a specific form unmarked.
It derives neither of the attested changes that extend a form up the hierarchy, a non-specific form becoming epistemic and an epistemic form unmarked.
The reversed hierarchy, with the non-specific layer on top, derives those two changes and loses the change from a specific unknown form to an epistemic one.
Coexpression as underspecification #
Each form of the four coexpression patterns is a single type. English some- imposes no restriction (AAA), Yakut -eme requires variation within an epistemic alternative and -ere constancy within one (ABB), Latin ali- requires variation across the alternatives and -dam constancy across them (AAB), and Russian -nibud', -to and koe- require variation within an alternative, variation across them and constancy across them, the entries (6) and (7).
The nanosyntactic lexicon gives Russian -to the profile of the specific unknown type, where the map makes it epistemic. The form shares the non-specific function with -nibud', and spellout, which gives a layer one form, hands that layer to -nibud' by the Elsewhere Principle.