Symbouletic modality
Igor Yanovich
MIT / Tübingen
CSSP September 27, 2013
symbouletic < συµβoυλυω‘advise’
Preview I: new modal flavor
Symbouletics added to [Portner, 2009]’s classification:
modality
kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk
SS SS SS SS SS SS SS SS SS SS SS SS SS SS SS S
dynamic priority
sssssssssssssssssssssssssssssssssssssss
2222 2222 2222 2222 2222 222
LL LL LL LL LL LL LL LL LL LL LL LL LL LL LL LL LL LL
L epistemic
(I can swim) (She might be here)
deontic teleological bouletic symbouletic
Igor Yanovich (MIT / Tübingen) Symbouletic modality 2 / 35
Introduction
Preview II: symbouletics among performatives
I use the framework for performatives by Condoravdi and Lauer, where:
Commissive: EP(SU,p) Directive: EP(SU,EP(y,p)) Imperative: EP(SU,p)
...and newsymbouletic: EP(SU, V
qbest(x,q))→EP(SU,EP(x,p))
Outline
1 Motivation for introducing symbouletic modality
2 Empirical properties of Russian symbouletic stoit
3 Formal analysis of stoit
4 Conclusion: symbouletics among other modals and other performatives
Igor Yanovich (MIT / Tübingen) Symbouletic modality 4 / 35
Motivating symbouletic modality
Known types of priority modality...
English priority -modals: must,have to,should,ought...
(1) Deontic, objective:
Tax office’s website: Everyoneshould file their taxes by April 15.
(2) Deontic, subjective:
Parent to a child: Youmustgo to bed at nine. (Because I set the rules.) (3) Teleological:
To get to the Polar Bear Park, youhave totake a plane.
(4) Bouletic:
Imusttry this cake. I simply must.
...and an unknown type
(5) You reallyshould go to that concert!
(6) Rough paraphrase of 5:
I advise you to go to that concert: it would be good for you to go!
5 is not (necessarily) deontic: it’s not about obligations 5 is not (necessarily) teleological: it’s not about goals 5 is not (necessarily) bouletic: it’s not about desires
Instead, 5 is about which course of practical action to choose
Igor Yanovich (MIT / Tübingen) Symbouletic modality 6 / 35
Motivating symbouletic modality
Symbouletic modality: the modality of advice
(7) Sue to Anna: You reallyshould go.
≈SueadvisedAnna to go.
(8) Sue to Anna: (According to the tax office,) youshouldfile by tomorrow.
6=SueadvisedAnna to file by tomorrow.
Matrix symbouletic claims may be described with advise.
Cf.: imperatives may be described withtell to;
performatives likeorderorpromise, with themselves.
Symbouletic modality is inherently performative
Regular priority modals describe the state of affairs (9) Ann: It’s you who has todo the dishes!
Mary: OKYou’re lying! (Today is your turn!)
Imperatives add a new to-do item
different formalizations: Portner, Kaufmann, Condoravdi & Lauer...
(10) Ann: Goto that concert!
Mary: #You’re lying! (That’s not what you tell me to do!)
Symbouletics pattern with imperatives: they are self-fulfilling (11) Ann: Youshould go to that concert!
Mary: #You’re lying! (That’s not what you are suggesting!)
Igor Yanovich (MIT / Tübingen) Symbouletic modality 8 / 35
Motivating symbouletic modality
Why didn’t we notice symbouletics before?
In English, shouldandought are core modals that can be used as symbouletics. But they are often used with “regular”, non-self-fulfilling modal semantics.
English (’d) better is close to a specialized symbouletic, but has not received much attention (but cf. [Denison and Cort, 2010]).
I am not aware of any studies of symbouletic uses ofbe worth (doing).
Today, we will look in detail at Russianstoit: it can be used as a symbouletic, but not as a neutral priority modal.
stoit: a modal specialized for advice
(12) OK Tebe you.dat
stoit stoit
poexatj go
v to
otpusk.
vacation
‘You should take a vacation.’
(13) * Soglasno according
pravilam, rules
tebe you.dat
stoit stoit
sdatj submit
otčot report
do before
zavtra.
tomorrow
‘According to the rules, you should submit the report before tomorrow.’
(14) Parent to child:Tebe you.dat
stoit stoit
pojti go
spatj.
sleep
OK if the parent mildly suggests the child better goes to sleep, but# if the parent intends to issue an order/describe an obligation.
advice deontic objective deontic subjective teleological
OK * * *
see 12 see 13 see 14 see 17
Igor Yanovich (MIT / Tübingen) Symbouletic modality 10 / 35
Empirical properties of symbouletic STOIT
stoit: a modal specialized for advice
(15) To improve her chances, Mashaoughtto buy a second lottery ticket.
But I wouldn’t advise that.
(16) OK Čtoby in.order.to
povysitj improve
svoi her
šansy, chances
Maše Masha.dat
nužno nužno
kupitj buy
vtoroj second loterejnyj
lottery
bilet.
ticket.
No But
ja I
by would
ej to.her
ne not
sovetoval.
advise (17) # Čtoby
in.order.to
povysitj improve
svoi her
šansy, chances
Maše Masha.dat
stoit stoit
kupitj buy
vtoroj second loterejnyj
lottery
bilet.
ticket.
No But
ja I
by would
ej to.her
ne not
sovetoval.
advise
16 withnužno is a neutral, disinterested teleological claim.
17 withstoit can only be advice, and needs to be personally endorsed by the speaker. The first sentence of 17 is fine. It is the continuation that is off.
Embeddability of stoit
Matrix modal claims with stoitalways constitute advice.
However,stoit can be embedded into sentencesnotproviding advice:
(18) Stoit stoit
li Q
mne I.dat
zapisatjsja register
na for
etot that
klass?
class
‘Should I register for that class (I wonder)?’
(19) Maša Masha
teperj now
dumajet, thinks
što that
Ane Anja
stoilo stoit.Past
tuda there
pojti.
go
‘Now Masha thinks that (according to Masha’s current information) it would have been better (given the circumstances back then) if Anya went there.’
Thus stoitis not a marker of advice. It just has a meaning which ensures that certain assertions with it necessarily become advice.
Similar to performatives:
‘I promise to go’ is performative, ‘I promised to go’ is not.
⇒Igor Yanovich (MIT / Tübingen) Symbouletic modality 12 / 35
Empirical properties of symbouletic STOIT
Our analytic strategy for stoit
Premise: stoit has a regular compositional meaning, which in certain circumstances gives rise to performativity.
⇒ common with [Condoravdi and Lauer, 2011], [Eckardt, 2012], etc.
Data collection: we identify the distributional properties ofstoit
Formal analysis: we formulate a compositional meaning that captures the identified properties.
We have already identified the property of Embeddability.
Properties of stoit: Decision
Letstoit(x, p) stand for “x stoitp”.
Decision: (I)x has the control over whether ap-future will actualize, and (II) it is not yet given thatp will actualize.
(20) Tebe you
stoit stoit
nanjatjsja get-employed
na prep
rabotu.
job
‘Youstoitget a job’
OK if it depends on the addressee to get a job: there are plenty of jobs around, she has relevant qualifications, etc.
#if there aren’t many jobs, it’s the employer’s market, and no qualifications would guarantee getting a place to work.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 14 / 35
Empirical properties of symbouletic STOIT
Properties of stoit: Decision
Decision II: it is not yet given that p will actualize.
(21) ContextBridge: one can get to the island only by crossing the bridge.
# Čtoby in.order.to
popastj get
na to
ostrov, island,
tebe you.dat
stoit stoit
pojti go
po on
mostu.
bridge
‘To get to the island, youstoitcross the bridge.’
Under embedding,Decisionbehaves as a presupposition:
‘Ann thinks Sue’s cello is expensive’ presupposes ‘Ann thinks Sue has a cello’
‘Ann thinks Suestoitgo’ presupposes ‘Ann thinks Sue has control over going, and it isn’t yet decided whether she’ll go.’
Properties of stoit: Subject Benefit
Subject benefit: to assertstoit(x,p) properly, the speaker must believe that acting towardsp is of direct benefit to x.
(22) Tebe you
stoit stoit
ispeč bake
pirog.
pie
‘Youstoitbake a pie’
#if the speaker wants to eat a pie, but there’s no direct benefit to the hearer in baking one.
OK if the hearer feels down, and baking a pie always lets him up.
Under attitudes it is the attitude bearer, not the speaker, to whom the belief inSubject Benefitis attributed.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 16 / 35
Empirical properties of symbouletic STOIT
Properties of stoit: Partial Rejection
With symbouleticshould in 11, the speaker cannot be accused of lying. For stoit, things seem a bit more complicated:
(23) Mary to Ann: Tebe you
stoit stoit
sxoditj go
na to
etot that
koncert.
concert
‘Youstoit go to that concert.’
a. Ann: # Ty you
lžoš:
lie ty you
ne not
predlagaeš suggest
mne me
tuda there
pojti.
go
‘You are lying, you are not suggesting that I go there.’
b. Ann: OK Ty you
ošibaješsja:
are.wrong mne I
ne neg
nravitsja like
etot that
dirižor.
conductor
‘You are wrong, I don’t like that conductor.’
c. Ann: ?Ty you
lžoš:
lie mne I
ne neg
nravitsja like
etot that
dirižor.
conductor
‘You are lying, I don’t like that conductor.’
Properties of stoit: Partial Rejection
stoit(x,p) introduces two claims: a self-verifying advice claim, and a Subject Benefitclaim.
The advice claim cannot be challenged.
But theSubject Benefit claim can.
Partial Rejection: (I) the act of assertingstoit(x,p) self-verifies that it is an act of providing advice, and cannot be challenged; but (II) that p is good for x is a regular assertion whose truth may be assessed in the usual manner
Igor Yanovich (MIT / Tübingen) Symbouletic modality 18 / 35
Empirical properties of symbouletic STOIT
Properties of stoit: Endorsement
Endorsement: the speaker has to endorse x’s acting towards p:
suggestions, orders, etc. to not do p directly contradictstoit(x,p).
(17) Čtoby in.order.to
povysitj improve
šansy, chances
Maše Masha.dat
stoit stoit
kupitj buy
vtoroj second
loterejnyj lottery
bilet.
ticket.
‘To improve her chances, Mashastoitbuy a second lottery ticket.’
# No But
ja I
by would
ej to.her
ne not
sovetoval.
advise
‘But I wouldn’t advise her to do that.’
(24) # Tebe you
stoit stoit
ispeč bake
pirog, pie
no but
ne not
delaj do
etogo.
that
‘Youstoitbake a pie, but don’t do that.’
Under attitudes,Endorsementshifts to the attitude’s subject.
Properties of stoit: Scope over Neg
Scope over Neg: whenstoitoccurs with a clausemate sentential negation, it always scopes over it:
(25) Context: The addressee has a choice of going to Boston, NYC or Philadelphia.
Tebe you.dat
ne not
stoit stoit
exatj go
v to
NYC NYC
=‘You shouldn’t go to NYC’ (>¬)
6=‘It’s not that going to NYC is your best option.’ (¬>)
But there is nothing wrong with a (¬>stoit) meaning as such:
(26) OK Eto this
ne not
značit, means
što that
tebe you.dat
stoit stoit
exatj go
v to
NYC, NYC
vedj as
v in
Bostone Boston
tože also interesno.
interesting
‘That does not mean you should go to NYC, because in Boston it’s also fun.’
Igor Yanovich (MIT / Tübingen) Symbouletic modality 20 / 35
Empirical properties of symbouletic STOIT
Properties of stoit: summary
Decision
Subject Benefit Partial Rejection Endorsement Scope over Neg
Embeddability: see 18 and 19
Schematic formal analysis of stoit
stoit(x,p):
presupposesdecision(x,p)
assertsbest(x,p)∧advise(SU,x,p),whereSU (fromSUggest-er) is the subject in a matrix context, and the attitude bearer under attitudes
decision(x,p) directly captures Decision: ‘among the possible futures at the evaluation index, there are somep-futures, and it depends on x’s actions whether ap-future or a non-p-future will materialize’.
best(x,p) stands for ‘p is best forx’.
More complicated stuff: advise(SU,x,p); temporal indexing.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 22 / 35
Formal analysis of symbouletic STOIT
Semantics of advice
I formalizeadvise in the framework of [Condoravdi and Lauer, 2012].
(27) advise(SU,x,p) :=
EP(SU,V
qbest(x,q))→EP(SU,EP(x,p)),
whereEP(a,q)stands for ‘ahas an effective preference forq’
Having aneffective preference(EP)forq means that your structure of preferences is such that qis one of your top priorities, and moreover, there are no conflicting priorities of the same rank. Thus if you are rational and have anEP forq, then you will act towardsq.
If youpublicly committedto someEP π, you created a correspondingPEP, and can now be accountable for acting as if you really effectively preferπ.
Semantics of advice
(27) advise(SU,x,p) :=
EP(SU,V
qbest(x,q))→EP(SU,EP(x,p))
V
qbest(x,q) is simply “all which is best for x”.
Thus EP(SU,V
qbest(x,q))essentially means “SU prefers what’s best forx, and moreover is going to act to achieve that”.
EP(SU,EP(x,p))= ‘SU has an EP for x to have an EP for p’
This is essentially the meaning of a directive.
It follows from it thatSU will act towards gettingx to act towardsp.
(27) = “ifSU effectively prefers what’s best for x, thenSU also effectively prefers gettingx to work towardsp”
Igor Yanovich (MIT / Tübingen) Symbouletic modality 24 / 35
Formal analysis of symbouletic STOIT
Semantics of advice
Paraphrasing our advice meaning:
“If your interests were the most important thing for me, I would have tried my best to get you do p”.
It follows that in the speaker’s opinion working towardsp is in the hearer’s best interests.
Introducing a conditional EP of her own, the speaker makes a public announcement that constrains her future actions. If it turns out x’s interests are an EP for her, she would be obliged to try to getx do p.
Advice among other EP -based meanings
Adapted from [Condoravdi and Lauer, 2011]:
(28) a. I promise that p. commissive
b. [[promise]]=λpλx.EP(x,p) c. [[(28a)]]=EP(sp,p)
d. Asserting (28a) makessp’sEP a public one,PEP. It thus restricts furthers actions bysp: ifsp doesn’t act towardsp, that makes the prior assertion of (28a) false.
(29) a. I order you to q. directive
b. [[order]]=λpλyλx.EP(x,EP(y,p)) c. [[(29a)]]=EP(sp,EP(hearer,q))
d. Asserting (29a) commitssp to anEP for the hearer to commit toq;
ifsp has authority to make orders to the hearer, that is enough to actually constitute an order.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 26 / 35
Formal analysis of symbouletic STOIT
Advice among other EP -based performative meanings
Commissive(promise): EP(SU,p) Directive(order): EP(SU,EP(y,p)) Imperative: EP(SU,p)
Symbouletic(advice): EP(SU,V
qbest(x,q))→EP(SU,EP(x,p))
Symbouletics are strictly weaker than directives, and thus may give rise to scalar inference: ifsymb(p), thendir(p)would have been too strong.
E.g. a parent saying “Youstoitgo to sleep” issues smth. less strong than “Go to sleep!”
But in case it is known thatEP(SU,V
qbest(x,q)), symbouletics and directives may collapse.
But if a parent clearly has anEP about what the child does (e.g. if the parent is glowering), “Youstoitgo to sleep” can have the same effect as an explicit directive.
Formal analysis accounts for stoit’s properties
Decision: encoded in the presupposition Subject Benefit: best(x,p)
Partial Rejection: best(x,p) is never self-verifying, but advise(SU,x,p) often is
Endorsement: advise(SU,x,p) contradicts explicit advice for ¬p Scope over Neg: not follows from the semantics, and should be listed as an idiosyncratic property of stoit
Embeddability: our meaning is embeddable
Igor Yanovich (MIT / Tübingen) Symbouletic modality 28 / 35
Formal analysis of symbouletic STOIT
Getting the temporal indices right
(19) Maša Masha
teperj now
dumajet, thinks
što that
Ane Anja
stoilo stoit.Past
tuda there
pojti.
go
‘Now Masha thinks that (according to Masha’s current information) it would have been better (given the circumstances back then) if Anya went there.’
Two temporal indices in 19:
tm, the higher index provided by the attitude (here, the present) tloc, the local index provided by the tense onstoit(here, the past) Adding correct indices to our meaning for stoit(x,p):
presupposesdecisiontloc(x,p) assertsbesttloc(x,p)∧
(EPtloc(SU,V
qbesttloc(x,q))→ ∃t0>tloc:EPtloc(SU,EPt0(x,p)))
Getting the temporal indices right
Assertive component of stoit:
λpλtlocλx.besttloc(x,p)∧
∧(EPtloc(SU,V
qbesttloc(x,q))→ ∃tt00>tloc :EPtloc(SU,EPt0(x,p)))
Assertion of (i) = “što Ane stoilo tuda pojti” from 19:
λwλtm.∃tloc<tm:besttloc(a,go)∧
∧(EPtloc(SU,V
qbesttloc(a,q))→ ∃tt00>tloc :EPtloc(SU,EPt0(a,go)))
Assertion of 19:
“Masha’s belief world-time pairs are a subset of the assertion of (i)[SU→m]”
Thus in Masha’s belief worldsw0 as they are in the presenttm, it was best for Anya to work towardspat past timetloc, and if Masha were to act in Anya’s best interests, she would have committed at pasttloc to getting Anya to form anEP forpat somet0.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 30 / 35
Conclusion
Relation of symbouletics to non-modal and modal cousins
The diachronic source of stoitis a verb meaning “to be worth”.
(cf. English “It’s worthq-ing” with the force of a suggestion.)
Semantic reanalysis went through the stage wherestoit(p) conveyed
“p is worth the effort”.
Gradually, suchstoitobtained the force of a suggestion.
The associated semantic change involved the introduction ofEP-s relative to the speaker.
Relation of symbouletics to non-modal and modal cousins
We can further reinterpret the meaning, introducing into it symbouletic modal baseSmbt(SU) :=V
q∃x :adviset(SU,x,q).
stoit’s assertion then becomesbesttloc(x,p)∧Smbtloc(SU)⊆p.
At some point, speakers might changeSmbtloc for another priority modal base, getting a less specialized modal. Just as other priority modals, Smbis relative to the local index (cf. [Hacquard, 2010]).
However, whether specialized symbouletics ever become general priority modals is unclear: more research is needed.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 32 / 35
Conclusion
Relation of symbouletics to non-modal and modal cousins
English ’d betterandbe worthare roughly parallel tostoit, but they are less embeddable when used symbouletically. As stoit, they arose from worth-related expressions.
English should andoughtacquired symbouletic uses after already being (semantic) modals.
⇒ many paths to symbouletic meanings
⇒ multiple ambiguity for symbouletics is not surprising
Conclusion
Symbouletic modality: new type of priority modality
Differs from other priority modals in having the same potential for performativity as performative verbs
Formally, symbouletic meanings do not have to be in the format of Kratzer’s modal semantics, but they can be so re-conceptualized.
Among performatives, symbouletics are strictly weaker than directives
This is formally captured by conditionalizing the internal directive EP(SU,EP(x,p)with a statement of the form “If I were to work in your best interests (but I’m not saying that I do), then...”
Igor Yanovich (MIT / Tübingen) Symbouletic modality 34 / 35
Acknowledgements
The termsymbouleticwas generously coined by Paul Kiparsky.
I am grateful for the discussions of various issues related to the present work to the audiences at UConn, Stanford and UCLA, and to Paul Bloomfield, Cleo Condoravdi, Kai von Fintel, Sabine Iatridou, Magdalena Kaufmann, Stefan Kaufmann, Paul Kiparsky, Manfred Križ, Sven Lauer, Yael Sharvit, Yakov Testelets, and Sam Wheeler.
The diachrony ofstoitwas studied using the Russian National Corpus, freely available at http://www.ruscorpora.ru/.
Condoravdi, C. and Lauer, S. (2011).
Performative verbs and performative acts.
In Reich, I., Horch, E., and Pauly, D., editors,Sinn and Bedeutung 15: Proceedings of the 2010 annual conference of the Gesellschaft für Semantik, pages 149–164, Saarbrücken. Universaar — Saarland University Press.
Condoravdi, C. and Lauer, S. (2012).
Imperatives: meaning and illocutionary force.
In Piñón, C., editor,Empirical Issues in Syntax and Semantics 9, pages 37–58.
Denison, D. and Cort, A. (2010).
Betteras a verb.
In Davidse, K., Vandenalotte, L., and Cuyckens, H., editors,Subjectification, intersubjectification and grammaticalization, pages 349–383. Mouton de Gruyter.
Eckardt, R. (2012).
Herebyexplained.
Ms., Göttingen University.
Hacquard, V. (2010).
On the event relativity of modal auxiliaries.
Natural Language Semantics, 18(1):79–114.
Portner, P. (2009).
Modality.
Oxford University Press.
Igor Yanovich (MIT / Tübingen) Symbouletic modality 35 / 35