Notes on Condoravdi 2002:

The Temporal Interpretation of Modals

1.0 Introduction

Claim: Non-root modals make a consistent temporal contribution

• non-uniformity follows from their interaction with the elements they combine with

Modals for the Present:

• eg. may, must, might, should, ought to
• have the perspective of the present
• future-orientation

Modals for the Past:

• eg. may have, must have, might have, should have, ought to have
• have perspective of the present
• past orientation

(ML: Even though these are morphologically perfect, they semantically seem past)

Adverbial Modification

• modals for the present: shift the EvalT forwards, sometimes, from UttT
◦ only compatible with present/future adverbials

• modals for the past: shift the EvalT backwards from UttT
◦ only compatible with past adverbials

• Question: Do modals contribute to the temporal interpretation

◦ directly?
◦ or through implicit tense?

• Condoravdi: Directly

1
 • What if modals are temporal as well as modal operators?

◦ modals would be associated with Reichenbachian temporal structures

• eg. forward-shifting, non-shifting, back-shifting modals

◦ they would quantify over worlds, as well as ∃ quantifying over times,

• eg. Priorian tense operators

• Condoravdi proposes a unitary semantics for all three cases

2.0 The Ambiguity of Modals for the Past

• (6a) is
◦ present-perspective, future-oriented (unambiguous)
◦ epistemic or metaphysical1

• (6b) is
◦ epistemic: present-perspective, past-oriented
◦ counterfactual: past-perspective, future-oriented

So “might have” is not, as one might expect, the perfect of “might”, but rather a future in the past

1 metaphysical is wrt. how the worlds may/might have turned out to be, re: live options

2
 • Why do modals of the past have this ambiguity?

◦ Karttunen (1972), G&Stokhoff (1975): Lexical Ambiguity

◦ Condoravdi: Scopal Ambiguity

3.0 Temporal Perspective and Orientation

3.1 Tense with Modals

• Q1: Do modals occur in the scope of tense?

◦ How do we determine the past/present perspective?

• Q2: Do modals combine with tensed or untensed sentences?

◦ How do we determine the EvalT, eg. orientation?

• Condoravdi: Modals
◦ A2: combine with untensed sentences
◦ A1: occur in the scope of tense

• Observation:
◦ modal auxillaries in unembedded clauses have only present-perspective

• Assumptions:
◦ there are zero tenses (cf. von Stechow 1995)
◦ Modals are in the scope of
▪ present tense (in extensional contexs)
▪ zero tense (in intensional contexts)

eg.

(Condoravdi will refer to this as “outer tense” - i.e., tense above a modal)

• Condoravdi proposes a decompositional analysis for modals for the past

▪ present-perspective: MODAL > PERFECT (epistemic)

▪ past-perspective: PERFECT > MODAL (metaphysical)

[Question: How do these two perspective options yield different modal bases? See section 4]

• Question: Do modals combine with tensed or tenseless sentences?

◦ How do we determine the EvalT?

• Morphologically, there's obviously no tense!

3
 • But what about semantically?

◦ McCawley (1971), Steedman (1997)

• Modals compose with tensed sentences

◦ Semantic present/future: morphologically bare

◦ Semantic past: morphologically perfect

▪ (ML: because infinitives can't take tense morphology?)

(Condoravdi will refer to this as “inner tense”, i.e., tense below a modal)

• The tenses PRES FUT PAST area deictic ones

◦ This can account for the adverbial pattern seen previously (in (1) and (2))

--> Condoravdi will argue that there is no tense in the scope of the modal

• The morphological perfect is actually a (semantic) perfect

• it interacts with the temporal contribution of the modal
• this will yield the ambiguity in (11a)

Condoravdi's Representation:

(11) a) He may be sick

a') MAY (he be sick)

b) He may have been sick

b') MAY (PERF (he be sick))

• Standard analyses of the perfect give it the back-shifting interpretation seen in (11b)

• The forward/non-shifting properties of ( 11a) will be dealt with later

◦ (ML: her proposal will be that modals extend forward the time of reference from t to [t,_) )

• Evidence that the morphological perfect is a perfect:

i) in some cases, it must be interpreted as a perfect, not a past

• eg. with the aspectual morphemes already and yet

4
 • already and yet are incompatible with eventive predicates
◦ yet they are compatible with “modals for the past”
▪ event with eventive VPs, but only with perfect morphology

◦ this suggests that the perfect morphology is a perfect, which allow

already and yet to combine with the sentence

(ML: I find almost all

of these good.
Although I very much
like the idea of the
perfect morphology
being a perfect.)

ii) if tense is in the scope of modals, we'd predict a different interpretation from what we
actually get in certain cases

• eg. strong intensional contexts

• Abusch (1997), von Stechow (1995):

◦ there can be no deictic tense in strong intensional contexts

◦ So if modal occurs in the scope of an intensional predicate

• we expect a zero tense, not PRES, FUT, or PAST

→ Yet we see shifting effects

(17) a) Mary believed that John may get sick (forward)

b) Mary believed that John may have been sick (backward)
c) Mary believed that John got sick (absent)
d) Mary believed that John was sick (absent)

• And so these shifting effects must come from something other than tense

• Condoravdi argues:
◦ forward-shifting comes from the modal
◦ backward-shifting comes from the perfect

eg. in counterfactual contexts

(18) a) He might have won

5
 b) PRES (PERF (MIGHT (FUT (he win))))
• A representation like (18b) predicts (18a) to be a future at UttT
◦ since non-zero tense operators are deictic),
◦ whereas it is actually a future in the past.

• Condoravdi concludes that the temporal localization of the sentence must be attributed to the modal
itself, and not an inner ten

• Question: What IS the temporal contribution of modals?

Generalisation:
• Modals for the present:
◦ stative/progressive predicates:
• optional future orientation
◦ eventive predicates:
• obligatory future orientation

3.2 The Semantics

• Assumed Ontology:
• Eventualities:
◦ events (eventive predicates)
◦ states (stative predicates)

• Temporal Intervals

◦ Verbal predicates take eventuality arguments

◦ Basic untensed sentences are predicates of eventualities

◦ intensional parameters (time and worlds) explicitly represented

◦ Tense operators
• map properties of eventualities to propositions

• instantiate these properties in time

◦ by locating e relative to a temporal interval ~ RT
• events: temporal inclusion
• states: temporal overlap

6
Temporal Operators rely on the AT relation:
Where AT(t,w,P) means that property P is instantiated in world w at time t
(19) AT (t,w,P)=
a) if P is eventive: ∃e[P(w)(e) & τ(e,w) ⊆ t
b) if P is stative: ∃e[P(w)(e) & τ(e,w) ∘ t
c) if P is temporal: P(w)(t)

(20) PRES: λP λw [AT(now, w, P)]

• This instantiates a property to now, a short interval

(21) PERF: λP λw λt ∃t' [t'≺ t & AT (t',w,P]

• This maps properties of eventualities (or times) to properties of times, and
• shifts EvalT of the element in its scope to an interval before the RT
◦ If PERF is directly under PRES, the RT is now
◦ if PERF is directly under a modal, the RT depends on the modal

• Modals are like the perfect

◦ the map properties of eventualities (or times) to properties of times, but
◦ they don't shift the EvalT
◦ they expand the EvalT forward (cf. Abney 1987)
◦ (and obviously they instantitate P in different possible worlds)
◦ [t,_) designates an interval with t as an initial subinterval, and extending to the end of time

(22) MAY/MIGHTMB: λP. λw. λt. ∃w' [w' ∈ MB(w,t) & AT ([t,_), w', P)]

(23) WOLLMB: λP. λw. λt. ∀w' [w' ∈ MB(w,t) ⟶ AT ([t,_), w', P)]

• As modals don't shift EvalT into the future, but rather expand the local EvalT into the future
i) default interpretation includes UttT
ii) PAST scoping over MODAL over PERF still yields present-perspective
(eg. givven what I will know)

7
 • Condoravdi treats already/yet/still as truth-conditionally trivial
◦ mapping properties of states or times onto themselves
◦ (being undefined if P is not remporal or stative)

(ML: I don't know how I feel about this. Then again, I have completely different judgements from what
she's provided regarding their acceptability)

• Does this yield what we want?

• Present perspective, present/future orientation

= modal + stative predicate
◦ Temporal overlap of the state with the interval [now,_)
• inclusion yields future orientation
• mere overlap yields present orientation
(as this allows the predicate to have held prior to now, as long as it extends at last
through to now)

• Present perspective, future orientation

= modal + eventive predicate
◦ Temporal inclusion of the event with the interval [now,_)
• yields future orientation (as above)

• Present perspective, past orientation (epistemic reading)

= modal + PERF (predicate)

(recall that PERF shifts the evalT to a time prior to the RT)

8
 ◦ Temporal inclusion of the event with an interval that precedes [now,_)
• yields past orientation

• Frame intervals (eg. yesterday) are intersective modifiers

◦ eg. yesterday takes a predicate of events, P, a world, w, and a time t,
and then states that t intersects with yesterday

• Modals in intensional contexts have their temporal perspective determined by the time associated
with the intensional context (eg. Hacquard's attitude time)

(17) a) Mary believed that John may get sick

b) Mary believed that John may have been sick
c) Mary believed that John got sick
d) Mary believed that John was sick

(17a) believe: introduces attitude time, t0, as RT

zero tense: ??
MOD: extends forward the RT to [t0,_)
get sick: eventive, so is included in the RT [t0,_)

→ This yields the forward-shifting effect

(ML: Recall that [ means that we don't include the endpoint)

(17b) believe: introduces attitude time, t0, as RT

zero tense: ???
MOD: extends forward RT to [t0,_)
PERF: locates the event wrt. an interval prior to [t0,_)

→ This yields the back-shifting effect (past orientation)

-whether the VP is predicative or stative (eg. been sick vs gotten sick)

(33) He might have won

a) PRES MOD PERF he win: present perspective, past-orientation
b) PRES PERF MOD he win: past perspective, future orientation

(33a) PRES: introduces now as RT

MOD: extends EvalT to [now,_) (in a possible world)
PERF: locates e in an interval prior to [now,_), t' (in a possible world)

9
 he win: eventive, so inclusion of event in t' (in a possible world)

→ According to my current evidence (at [now,_) ) PRESENT PERSPECTIVE

in a possible world, he won prior to [now,_) PAST ORIENTATION

(33b) PRES: introduces now as RT

PERF: locates EvalT (of modal) to an interval prior to now, t'
MOD: extends time of winning to [t',_) (in a possible world)
he win: eventive, so inclusion of e in [t',_) (in a possible world)

→ Given the circumstances at t' (which is prior to now) PAST PERSPECTIVE

in a possible world, he won in a time interval [t',_),
where t' precedes now FUTURE ORIENTATION

→ NOTE: This doesn't yet get us the counterfactual nature of (33b)

4.0 The Modal Base

• Metaphysical vs Epistemic Modal Bases

◦ Epistemic: related to what the speaker takes the actual world to be now

◦ Metaphysical: related to how the world may turn out to be, given that at present, there
are a number of live options

These are difficult to disentangle, as if an issue, p, is taken to be not settled, then the metaphysical
live options are also the epistemic alternatives.

• Settledness/Historical Necessity
• “A sentence is historically necessary at a time t, if it is true at t, regardless of what the
future is like.”

• Thomason (1985): World-Time Model

▪ TxW frames

▪ the set of times is a linear structure

▪ 3-place relation TxWxW, ≃

i) for all t, ≃t is an equivalence relation, and

ii) for any world, w, w' ∈W and t', t' ∈ T

if w1≃t w2 and t' ≺ t,

then w1 ≃t' w2

10
(ML: So basically, if w1 and w2 are equivalent at t, and t' precedes t, then w1 and w2 were equivalent at
the previous time, t'. So possible worlds can only branch off in the future. You can't have convergence of
different possible worlds)

• Generalisation:
◦ A non-root possibility modal has an exclusively epistemic reading when the instantiation of the
property it applies to is presupposed to be historically necessary if true.

(44) Settledness for P:

For any world w' ∈ CG, and any w'' such that w' ≃t0 w'':
AT([t0,_), w', P) iff AT ([t0,_), w'', P)

(ML: So as long as P holds for the interval [t0. _) for each world. Does this mean that it can have taken
place at different subintervals in [t0,_)

• A non-root possibility modal can have a metaphysical modal base only if the common ground
of the context does not satisfy settledness for P

• (i.e., it must satisfy Diversity (see paper for definition, it's basically non-settledness)

• So what about the counterfactuality?

Consider the temporal interpretation of “He might have won”

past-perspective, future-orientation

i) The past-perspective says that P was a live option at t', w here t' ≺ now

ii) There are more live options at t' than there are worlds in the CG at now,
(because we are assuming worlds branch off, but don't converge)

By asserting that P is a live option at t',

where we have a larger number of possible worlds than in the CG at now,
we implicate that P is not verifiable by the CG at now

→ This yields the counterfactual reading,

as the past possibility is implicated to be an unactualized one.

11