Non-Culminating accomplishments in Mandarin

LicensingNon-Culminating
accomplishmentsinMandarin.
Experimental&theoretical
evidence.
TELIC2017,UniversitätStu7gart,12-14/01/2017
HamidaDemirdache(LLINGNantes/CNRS)
JinhongLiu(LLINGNantes/CNRS)
FabienneMarMn(U.Stu7gart)
HongyuanSun(LLINGNantes/CNRS,CERCLL/U.dePicardieJules-Verne)
1
Introduction
MandarinChineseisreportedtobealanguagewheretransiMvechange
of state (CoS) verbs license non-culminaMng (NC) readings. (Tai 1984,
Chief2007,Koenig&Chief2008,Demirdache&MarMn2015)
• Li7leworkonlicensingofNCreadingsinMandarin
“Zero-CoSreading”(Tatevosov&Ivanov’s2009failed-a7empt)
“ParMal-CoSreading”(Tatevosov&Ivanov’s2009parMalresult)
• ContribuMonsofthisstudy:
• Relevantverbclasses
• Experimentalevidence(Liuinprep.)ontheroleofiteraMve
adverbsinlicensingNCreadings
• ContribuMonoftheaspectualmarkerle
2
Introduction:Mandarinaccomplishments
•  2typesofaccomplishmentsinMandarin:
-Monomorphemicverbs(MMVs)
-ResultaMveverbcompounds(RVCs):
• MostMandarinaccomplishmentsareRVCs:
acMvity(V1)+resultaMvecomplement(V2).
(Li&Thompson1981,Tai1984,Sybesma1999,2013,Lin2004)
ex.cā-diào‘wipe-drop’:wipe/erase,dǎ-suì‘hit-break’:break
•  ThisstudyfocusesonMMVs
•  Verbsinlimitednumber(Lin2004:53);
•  MMVsallownon-culminaMngreadings;
•  OvertresultaMvecomplementstriggereventculminaMon.
3
Non-culminatingCoSverbsinMandarin
Roadmap
Part1.Experimentalevidence
Mandarin follow up on the experiment just presented by Angeliek with
iteraMveadverbs→robustevidenceforNCCoSconstruals&theACH
Part2.Theore3caldiscussion
DisMnguish2classesofMMverbs:dependingonwhethertheyrequireor
not1requireanadverbialtolicensezero-CoS
•  Whatisthesourceofthenon-culminaMngreadings?
•  Eventstructure&lexicalsemanMcsofCoS-MMVs
→ NotacMvites,NotcoercedintoacMvitypredicatesonzero-CoSreading
→ Donotinvolveacoverttry-head
•  Thesourceofthenon-culminaMngreadings:verballe
Altshuler2014:Hindisimpleverb-perfecMveasaparMMveoperator
DerivingZeroCoSconstruals
Furtherevidence:Boundednessrequirement
4
IteraMveAdverbsIncreaseZero—CoS
ReadingofMMV
ExperimentalEvidence
5
1st Experiment:
Testing the Agent Control Hypothesis with non-culminating events in Mandarin
(2016 DGF workshop, Angeliek van Hout’s Telic 2017 talk for crosslinguistic
comparison)
◈ 
Participants
◈ 
◈ 
30Mandarinnativespeakers
Fullvs.ZeroChangeofState
◈  TruthValueJudgmentTask
◈ 
◈ 
Agentsvs.Causers
8MMCoS:
suì(break)
kāi(open)
zhé(cut)
guān(close) mái(bury) shā(kill)
jiě(untie)
xī(blowout)
TestingProbes:YesorNo
Agent-Subject
Cause-Subject
(MMV-PERF)
1.  Hǎidào
guān-le
nà
Pirate
close-PERF
that
‘Did the pirate close that door?’
2.
shàn
CLF
mén ma?
door Int?
Nà-zhèn feng
guān-le nà shàn mén ma?
That –CLF wind
close-PERF that CL F door Int?
“Did the wind close that door?”
Results: MMV
DfSumSqMeanSqFvaluePr(>F)
SubjectType18687868727.736.51e-07***
Situation1144838144838462.26<2e-16***
Subjecttype:situation13979397912.700.000532***
Results: Mandarin MMV
•  Mandarin
simple verbs:
Participants
accepted zero-CoS significantly more
often
for Agent than for Causer subjects (F=27.73, p<.001).
•  Confirms the role of agenthood, as predicted by the
ACH, with culmination behaving as a cancellable
implicature with Agents, but as an entailment with
Causers.
● 
● 
● 
Questions
Why is there only 38 % of acceptance for the
nonculminating reading in Agent-Zero CoS
condition?
Sun’s observation: ZeroCoS reading with certain
verbs is in fact acceptable only when the verb is
modidied by an adverbial, like haoji-ci `several
times’.
Can adding an iterative adverbial increase the
acceptance of nonculminating readings in AgentZeroCoScondition?(2ndexperiment)
2nd Experiment:
MMV + hǎojǐcì “several times”
◈ 
Participants
◈ 
◈ 
20Mandarinnativespeakers
ZeroChangeofState
ShortmovieclipsshowingeventswithnosuchCoSatall(as
encodedbythepredicate
◈  TruthValueJudgmentTask
◈ 
Agentsvs.Causers
SubjectargumenteitheranAgent(clown,pirate)
oraCauser(wind,explosion)
DESIGN
◈ 
2x1designvaryingSubjecttype
Agent
PureCauser
ZeroCoS
8
8
16testingprobes(8*2)
2typesofpurecausers:wind,explosion
8MMV:
◈ 
◈ 
◈ 
suì(break)
kāi(open)
zhé(cut)
guān(close) mái(bury) shā(kill)
jiě(untie)
xī(blowout)
TVJ Task
1
WatchingshortMovieClip(NoCoS)
2
TestingProbe:YesorNo
Agent-Subject
Cause-Subject
3.  Nà-gè hǎidào guān-le
hǎojǐcì
That –CLF pirate close-PERF several times
“Did the pirate close that door several times?”
4.  Nà-zhèn feng
guān-le
hǎojǐcì
That –CLF wind close-PERF several times
“Did the wind closed that door several times?”
MMV-PERF+iterative adverb
nà-shàn mén ma?
that CLF door Int?
nà-shàn mén
that CLF door
ma?
Int?
Materials & Design
◈ 
Agent-ZeroCoS(close)
◈ 
“Did the pirate close
that door several
times?”
Cause-ZeroCoS(close)
“Did the wind close
that window several
times?”
Results
MMV + Iterative Adverb
Table 1. Descriptive Statistics
Condition
Agent-Zero
CoS
CauseZero CoS
Mean
percentage of
“yes”
responses
0.82
Number of
“yes”
responses
Std.
Deviation
121 (147)
0.12750
0.05
8 (150)
0.12752
Results: MMV+Iterative Adverb
Result 1: Agenthood
Participants accepted zero-CoS significantly more
often for Agent than for Causer subjects
(F=382.932, p<.001).
✓ ( Re)confirms the role of agenthood, with
culmination behaving as a cancellable implicature
with Agents, but as an entailment with Causers.
Agent Zero Vs. Cause-Zero
◈ 
Figure3:Adultyesresponsesacrossverbtypes
◈ 
ThesubjectofthesethreeMMV+iterativeadverbiswind.
Agent Zero Vs. Cause Zero
20adults:82%Yesforagent-zero
◈  2typesofCausers
◈  Explosion(4items):break,cut,bury,kill
◈  Wind
4items):open,close,unMe,blowout
◈  Only3adults:say“yes”onthecause-zerocondiMon
◈  8/9yesresponsesfor3/4items:open,un3e,close
wherethecauseristhewind.
◈  PersonificaMonofthewind?
◈ 
◈  Noyeswith‘blowout?Accidentorsomethingtoexplain?
Comparing Results across
experiments
◈ 
Exp1Agentzero(withoutadverb):38%“Yes”Responses
◈ 
Exp2Agentzero(withadverb):82%“YesResponses
◈ 
Exp1Causezero(withoutadverb):7%“Yes”Responses
◈ 
Exp2Causetzero(withadverb):5%“Yes”Responses
◈ 
Figure4:%of“yes”acrossverbclassesin
1stvs.2ndexperiments
Result 2 Iterative Adverbs
Chineseadults(andchildren)acceptmore
easilyAgentzeroCoswithanadverbsuch
asseveraltimes.
◈  Condirmstheroleofiterativeadverbsin
facilitatingzeroCoSconstruals.
◈ 
Theoreticaldiscussion
23
2subclassesofMMverbs
RobustexperimentalevidenceconfirmingD&M’s(2015)
claim that the zero-CoS reading is possible with an
agenMve subject, but not with a non-agen3ve subject,
andthisevenwhentheverbismodifiedbyanitera4ve
adverb.
→ DisMnguish2subclassesofverbs:
WithanagenMvesubject
• Class1:zero-CoSOKwithoutaniteraMveadverbial
• Class2:zero-CoSoutwithoutaniteraMveadverbial
Whyisthezero-CoSreadingacceptablewithoutanadverbialwithclass
1verbsbutonlywithanadverbialwithclass2verbs?
24
CoSMMVs:2subclasses
Class1(larger):shāo‘burn’,rǎn‘dye’,sī‘tear’(cf.table1)
•  ParMalCoSreading:→OK
TheCoSdoesnotoccurtoanyposiMvedegree
•  ZeroCoSreading: →OK
AproperpartofthelexicalizedCoSoccursonly
(1)Yuēhànshāoletā-deshū,dànméishāo-zháo/shāo-huǐ
YuehanburnPERF3SG-DEbookbutNEGburn-touch/burn-destroy
‘Yuehanburnedhisbook,butitdidn'tgetburnedatall/completely.
(2)Tāsīlenèi-geběnzi, kěshì(běnzi
3SGtearPERFthat-CLnotebook
butnotebook tàihòu)méi(wánquán) sī-huài
toothickNEG(completely)tear-damage
‘Shetorethatnotebook,butdidn’t(completely)tearitup(thenotebook
beingtoothick).’
25
CoSMMVs:2subclasses
Class2(smaller):shā`kill’,chú‘getridof(atyrant)’,zhāi'pick(aflower)'.
•  ParMalCoSreading→No
•  ZeroCoSreading→No
(3)#TāshāleYuēhàn,
Yuēhànhái
huó
zhe.
3SGkillPERFYuehan YuehansMll
alive
DUR
Intended:‘HekilledYuehan,butYuehanissMllalive.’
(4)#Qùnián,cūnmínmen chú
le
cūnzi
lǐ
nèi-ge
last.yearvillagersget.rid.of PERF
village inside that-CL
èbà,
èbà
hái
zài.
tyrant,
tyrant sMll
exist
Intended:‘Lastyear,thevillagersgotridofthelocaltyrant,butthetyrantis
sMllthere.’
[exceptinasituaMonwherethetyrantcamebackaŠerbeingexpelled.]
Notes:JudgementvariaMon:shā`kill’allowszero-Cosreading(Talmy2000)
26
MonomorphemicCoS:2subclasses
Table1
Class1
shāo'burn',
sī'tear',
rān'dye'
mái‘burry’
fā‘leaven’,
zhé‘cut’,
jiě‘unknot(a3e)’,
Class2
shā'kill',
chú'getridof(thetyrant)'
zhāi'pick(aflower)'
xī‘blowout’
suì‘break’
ParMalCoSreading
√
#
ZeroCoSreading
√
#
√
√
#
#
for-adverbial
√
√
in-adverbial
√
√
ZeroCoSreadingwithonce/
severalMme(s)
ZeroCoSreadingwithcauser
subject(+once/severalMmes)
27
AdverbsthatlicenseNCreadings
Iterativeadverbs
Thezero-CoSreadingofverbsofclass2isacceptableonlywhentheverbis
modifiedbyaniteraMveadverbial(Demirdache&Sun2014),likehǎojǐ-cì
`severalMmes'(5),seealsoTai(1984:291).
(5)Tā shāle Yuēhànhǎojǐcì, Yuēhànhái huó
zhe
3SGkillPERF Yuehanseveral.MmesYuehansMll live
DUR
Literally:‘HekilledYuehanseveralMmes,butYuehanissMllalive.'
Liu(inprep.)observesthat(5)isalsosalvagedwithyí-cì‘once'(6).
(6)TāshāleYuēhànyícì,Yuēhànháihuó
3SGkillPERFYuehanonceYuehansMlllive
Literally:‘HekilledYuehanonce,butYuehanissMllalive.’
zhe
DUR
DeSwart(1991):adv.likeonce/several4mesareassociatedwithaplurality
condi3ononquan3fyingdomains(thatforbidsquanMfyingonasetwhose
cardinalityisknowntobelessthantwo).
(5)and(6)normallydonottriggertheoddinferencethatdeathisreversible:
theadv.quanMfiesoverunsuccessfula7empts.
28
AdverbsthatlicenseNCreadings:
Durativeadverbs
DuraMveadverbialslikeshífēngzhōng‘fortenminutes’(7)orliǎng-nián
‘fortwoyears’(8)alsolicensezero-CoSreadingsofverbsofclass2.
(7)Nóngfūshā nèi-tóuniúshāle
shífēngzhōng
farmerkill that-CLoxkillPERF tenminute
niúdouméisǐ
oxDOUNEGdie
‘Thefarmerkilledtheoxfortenminutes,buttheoxdidn’tdie.’
(8)Cūnmínmen chú
nèi-ge èbà
chú
leliǎng-nián,
villagersget.rid.of that-CL tyrantget.rid.ofPERFtwo-year
èbà hái
zài.
tyrant
sMll
exist
Literally:‘Thevillagersgotridofthetyrantfortwoyears,thetyrantissMll
there.’
29
EventstructureandlexicalsemanticsofCoSMMVs
Notactivityverbs
MandarinCosMMVsareac3vityverbsconven3onallyassociatedtoaresultof
acertaintype(Talmy2000,Chen2016)(e.g.wash-verbsinEnglish).
→Theresultisimplied(bythecontext),ratherthanentailed/encodedbythe
verb.CancellaMonisexpected.
Counterargument:Verbsofclass1/2acceptboththeso-calledcounterfactual
andscalarreadingsofchàdiǎn`almost’,see(9).
(9)a.Lùlu
chàdiǎnshāo leyì-běnshū
Lulu
almostburn PERFone-CLbook
‘Lulualmostburnedabook.’
[Luludidn’tputitintofire.]OR[Luluburnedthebook,butnotthewholebook.]
b.Nóngfū chàdiǎnshālenèi-tóuniú
farmer
almostkillPERFthat-CLox
‘Thefarmeralmostkilledtheox.’
[The farmer chose another ox aŠer hesitaMon.] OR [The ox survived from an
eventthatcouldleadtoitsdeath.]
Conclusion:verbsofclass1/2arecausa3veaccomplishments.
30
EventstructureandlexicalsemanticsofCoSMMVs
Notcoercedintoanactivityverb
Verb+adverbial>coercedintoanacMvity?
‘Subtrac3ve coercion’ hypothesis: CoS MMVs allow the zero-CoS reading iff
theyarereinterpreted(throughcoercion)intoacMvitypredicates.
→ThecoercedverbdoesnotentailaCoSanymore(inBo7’s2010terms,they
are`subtracted’oftheirculminaMonpoint).
CounterArgument:
AccomplishmentsmodifiedbyoncearesMllaccomplishments
Pinon(2005):denyingculminaMonispossibleonlywithtelicpredicates.
à“(not)completely”(adv.ofcompleMon)isoddwithatelicpredicates
(10)Heatehiscereals,butnotcompletely.
(11)Heatecereals,#butnotcompletely.
Accomplishmentverbkeepsitsaccomplishmenthood,evenwhenmodifiedby
‘possiblycoercing’adverbials,suchas“fortenminutes”(12).
(12)Heatehiscerealsfortenminutes,althoughnotcompletely.
31
EventstructureandlexicalsemanticsofCoSMMVs
Notcoercedintoanactivityverb
MandarinaccomplishmentMMVskeeptheiraccomplishmenthoodevenwhen
modifiedbyadverbialslikeyícì‘once’,see(13)below:
(13)Lùlushāoguo nèi-xiē shùyè(yícì),suīránméiquán shāo
LuluburnEXP that-Cl.PlleafoncealthoughNEGcomplete
burn
‘Luluburnedthoseleaves(once),althoughnotcompletely.’
(14)Lùlu shāoguo (yícì)
shùyè, #suīránméi
quán
shāo
Lulu burnEXPonce
leaf
althoughNEG
completeburn
‘Luluburnedleaves(once),#althoughnotcompletely.’
Conclusion:verbsofclass1/2arenotcoercedintoac3vi3es
32
EventstructureandlexicalsemanticsofCoSMMVs
Nocoverttry-head
AlternaMveaccount:a(silent)voiceheadmeaning‘try’
shā‘kill’=chángshì[try]shā
Thezero-CoSreadingcomesfromthefactthattryVdoesnotentailV.
•  PerfecMve sentences with causaMve change of state verbs entail that the
causaMon event started (i.e. a causing acMon of the relevant type must
havestarted),evenunderthezero-CoSreading;
•  Thetry-counterpartofthesesentencesdoesnotcarrythisentailment;that
is, it is compaMble with a situaMon where the agent has not started to
performacausaMoneventoftherelevanttype.(SeeGrano2011)
33
EventstructureandlexicalsemanticsofCoSMMVs
Nocoverttry-head
InasituaMonwhereLuluisunknowinglyparalysedinherbedandonlymentally
triedtokillacockroach(withoutmanagingtodoanymovement),(15)belowis
false,butitsovertlyconaMvecounterpart(16)istrue.
(15)Lùlushā le
yí-cì
nèi-zhīzhāngláng
Lulukill
PERFone-Mme
that-CLcockroach
‘Lulukilledthecockroachonce.’
(16)Lùlushì
zhe
shāleyí-cì
nèi-zhīzhāngláng
Lulutry
DUR
killPERFone-Mmethat-CLcockroach
‘Lulutriedtokillthecockroachonce.’
Zero-CoS reading of accomplishments require more than a try (but less
thanasuccess),cf.MarMn2015
--nocovertconaMveheadinvolvedintheirsemanMcs.
--TheEnglishtryisamisleadingtranslaMonofCoSMMVsunderthezeroCoSreading.
34
Thesourceofnon-culminatingreadings:
theaspectualmarkerle
AspectualmarkerleinMandarinismostcommonlyreferredtoasaperfecMve
marker (Wang 1965; Chao 1968; Smith 1991, Sybesma 1999, Lin 2006, Sun
2014,a.o.),
ManyauthorsdisMnguish
•  verballe:perfecMve
•  sentencefinalle:inchoaMve-->“currentlyrelevantstate”
(Li&Thompson1981:238,Paul2015)
Concernedhereonlywithverbal-le,whichwetaketobeaperfecMvemarker.
DrawingaparallelbetweenMMV-leandtheHindiSimpleVerb-perfecMve
(SV-PFVHI)onAltshuler’s2014account,
→  arguethatthesourceofNCconstrualsinMandarinisMMV-le
→  Advantage:crosslinguisMcdifferencebetweenMandarin&Englishdoesnot
lieinthelexicalmeaningofCoSverbs(e.g.ofburnorkill)acrosslanguages.
35
Altshuler2014:
SimpleVerb-perfectiveinHindi(SV-PFVHI)
asapartitiveoperator
SV-PFVHI is a parMMve operator, combining with a VP and
requiringthattherebeanevente'intheworldofevaluaMon
w*thatisastageofaVP-eventeina‘nearenough’worldw.
DiffersfromtheProgressiveinonecorerespect:
(17)Progrequiresthate’beapropersubpartofe(e’⊂e)
(18) SV-PFVHImerelyrequiresthate’beasubpartofe(e’⊆e)
e’=e→culminaMngCoSreading
e’⊂e→nonculminaMngCoSreading
Applied to an accomplishment VP does not lead to a
culminaMonentailmentassumingthataccomplishmentevents
haveatlasttwostages.
36
DerivingzeroCoSconstruals
AssumingthatverballeisaparMMveoperator requiringthate’bea
subpartofe(e’⊆e)
→  Zero CoS construal arises when e’ is an event part that
excludesanychangeofstate.
→CorrectlypredictsthatperfecMvesentenceswithCoSverbs:
i. 
require more than a try: entail that the causaMon event started
(that a proper causing acMon of the relevant type must have
started)evenunderthezero-CoSreading[Seediscussionof(15)vs
(16)Nocoverttry-headslide]
ii.  ButevenaveryminimaliniMalpropereventpartsufficestomakea
telic LE sentence true: e.g. (19) will be true as soon as the hair
dresserstartsapplyingthedyeonthehair.
(19)Fàxíngshī rǎn le
tā
de
tóufa
hair.dresserdye PERF 3SG
DE
hair
‘Thehair-dresserdyedherhair.’
37
Furtherevidence:
Theboundednessrequirement
AlthoughSV-PFV
HIallowsNCconstruals(20),itisincompaMblewiththeProg(21):
(20)maĩneaajapnaakekkhaayaa,(aurbaakii
kalkhaũũgaa).
I.ERGtodayminecakeeat.PFVandremainingtomorroweat.FUT
‘Iatemycaketoday,(andIwilleattheremainingparttomorrow).’(Singh91)
(21)maayaa-ne biskuT-kokhaa-yaa#aur useabtakkhaa rahiihai
May-ERG
cookie-ACCeat-PFVand itsMlleatPROG bePRS
Intended:`Mayaateacookie,andissMlleaMngit’(Altshuler2014:759)
→Why?BecauseSV-PFVHIimposesaboundednessrequirement:e’mustbeabounded
eventpart(didnotdevelopfurtherintheworldofevaluaMon,possiblybecauseitwas
completed)
→ 
Samecontrasta7estedwithMandarinLE:
(22)Lùlukāi-lenà-shànmén,dànshìméngēnběnméikāi.
Luluopen-PERFthat-CLdoorbutdooratallnotopen
`Luluopenedthedoor,butitdidn’tmoveatall.’
(23)Lùlukāi-lenà-shànmén,#érqiěháizàikai
Luluopen-PERFthat-CLdoorandsMllPROGopen
`Luluopenedthatdoor,andsheissMllopeningit.’
38
Conclusion
Q.  Whatisthesourceofnon-culminaMngCoSreadingsinMandarin?
TheperfecMvemarker:verbalLE
Welcomeimplica4on:
MandarinClass1/2verbsarestandardcausa3veverbs
NodifferencesinmeaningbetweenMandarinCoSverbssuchaskillor
burnandtheirEnglishcounterparts
Proposal explains why Mandarin sentences with perfecMve CoS verbs
require more than a try, but even a very minimal ini4al event part
sufficestomakethemtrue
Remainingques4on:
WhydoiteraMve/duraMveadvsincreasetheacceptabilityofNCCoS?
Rough illustraMon: by e.g. coercing atomic events (‘pick’) into events
withstages/eventparts.Tobecon4nued.
39