Event-Relative Modality #
@cite{hacquard-2010} @cite{kratzer-1981} @cite{cinque-1999} @cite{cinque-2004} @cite{rizzi-1978}
Part I: Italian Restructuring #
Connects the Italian restructuring data (Fragments/Italian/Modals) to @cite{hacquard-2010}'s content licensing theory (EventRelativity §8).
The Argument #
- Italian potere/dovere can restructure (appear below AspP).
- When restructured, these modals lose epistemic readings.
- Content licensing explains WHY: restructured modals are bound to the VP event (by aspect), and VP events lack propositional content, so epistemic modal bases cannot be projected.
- When non-restructured (above AspP), the modal binds to the speech event (or attitude event), which IS contentful → epistemic available.
This is the key empirical argument for event-relative modality: the same lexical modal (potere) shows different flavor availability depending purely on its syntactic position, explained by content licensing.
Part II: Event Projection and Unattested Pairs #
@cite{hacquard-2010}, §4.2: modals are keyed to (individual, time) pairs, but not all combinations are attested. A modal must be keyed to the participants and running time of the MOST LOCAL event. Event projection (holder(e), τ(e)) derives the correct pair for each event binder, explaining why certain pairs are systematically absent.
Restructuring forces the modal below AspP. This maps to
ModalPosition.belowAsp in the EventRelativity framework.
Equations
Instances For
Non-restructured modals sit above AspP.
Equations
Instances For
Content licensing predicts that restructured (low) modals cannot be epistemic: they are bound to the VP event, which lacks content.
This single theorem explains ALL the restructuring data:
- potere_high: epistemic ✓ because high modal → speech act → content
- potere_low_clitic: epistemic ✗ because low modal → VP event → no content
- dovere_high: epistemic ✓ (same reasoning)
- dovere_low_aux: epistemic ✗ (same reasoning)
The empirical data matches the theoretical prediction for potere.
The high/low flavor sets in the fragment entry align with the event binder's available flavors at each position.
Same bridge for dovere.
Both potere and dovere are single lexical items: the same verb appears high (with epistemic) and low (without epistemic). This rules out lexical ambiguity as an explanation — the flavor restriction follows from structural position alone.
@cite{hacquard-2010}, §1: Italian potere and dovere express both epistemic and root modality with the same lexical item, and the availability of epistemic readings tracks the syntactic position.
If epistemic/root were lexically distinct modals (as in some analyses of English can_epis vs can_root), we would expect no syntactic correlation. But Italian shows that ONE lexical item exhibits the restriction purely based on position. Content licensing explains this without positing ambiguity.
Furthermore, the restriction is PRODUCTIVE: any restructuring modal loses epistemic in the restructured position. The theory predicts this for ALL restructuring modals — it's not a per-item stipulation.
@cite{hacquard-2010}, §4.2: modals are keyed to (individual, time) pairs, but not all combinations of individuals and times are attested.
| Individual | Time | Attested? | Example |
|---|---|---|---|
| speaker | speech time | ✓ | epistemic have to |
| attitude holder | attitude time | ✓ | embedded epistemic |
| VP participant | VP time | ✓ | root have to |
| speaker | VP time | ✗ | — |
| VP participant | speech time | ✗ | — |
The missing diagonal pairs (speaker + VP time, subject + speech time) are explained by event projection: each event binder projects a FIXED (individual, time) pair. There is no event that pairs the speaker with the VP time, or the subject with the speech time.
The three event binders each project a specific (individual, time) pair. This is why not all combinations are attested — pairs not projected by any event are systematically absent.
@cite{hacquard-2010}, §4.2: "a modal seems to be relative to an individual and a time, but not all time/individual pairs are attested. Instead, the modal has to be keyed to the participants and running time of the most local event."
The paper's central claim: the position → flavor correlation is DERIVED from content licensing, not stipulated.
High modals (above AspP) bind to contentful events → epistemic available. Low modals (below AspP) bind by aspect to the VP event → no content → no epistemic. This dissolves @cite{cinque-1999}'s puzzle without dedicated functional heads for each modal flavor.
@cite{hacquard-2010}, §6.3: "high modals tend to be epistemic and low modals circumstantial, without having to stipulate two separate entries for each modal."
Events carry propositional content that (individual, time) pairs
do not. This is the key advantage of event-relative modality over
pair-relative modality: the content licensing predicate hasContent
discriminates events even when they project to similar pairs.
@cite{hacquard-2010}, §6: "what sets speech and attitude events apart from ordinary events is (what I am calling) their associated propositional 'content', which I take to be crucial for licensing epistemic modal bases."