1

1/81

Findingtenseswherewedon’texpectthem

LisaMatthewson

Tenselessness WorkshopCentre forApplied Research andOutreach in LanguageEducation

University ofGreenwichOctober 6,2017

2/81

Thequestion

? Howshould weanalyze superficially tenseless languages?

• As tenseless?(Bohnemeyer2002,2009,Shaer2003,Wiltschko2003,Bittner2005,2008,2014,Lin2006,2010,2012,Tonhauser2011,Mucha 2013,Ritter&Wiltschko2014,…)

• Oras tensed?(Matthewson2005,2006, Jóhannsdóttir &Matthewson2007,ReisSilva&Matthewson2007,Sybesma 2007,Hayashi2011,Thomas2012,Sun2014,…)

• Wecan always postulate null tenses – evencompletely null tenseparadigms. But should we?

3/81

Thechallengewhendeciding

• Everyanalysis hasawayofdealing with the factthat temporal reference isnot random or radically ambiguous in superficially tenseless languages.

Bohnemeyer (2009):

• ‘Topic times playa role in the interpretation ofutterances whether ornotthese are tensed, and theprinciples involved in their contextual resolutionare thesame in tensed and tenseless languages.’

Tonhauser (2015):

• ‘temporal reference in tenseless languages under tenseless analyses is justas specific as in tensed analyses of tensed (and tenseless) languages…because such analyses allow foravarietyof factors (context,adverbials,grammatical aspect, etc.)contributing to temporal reference.’

4/81

Conceptualvs.empiricalarguments

• Different authors havedifferent conceptual preferences.

• e.g.,Tonhauser (2015) invokes Occam’sRazor toargue against null tenses,while others maydisfavour cross-linguistic variation inpragmatic principles.

• Empirical arguments are theholygrail, but finding knock-down argumentsforcovertelements is challenging.

5/81

Drawingevidencefromarangeoflanguages

• Oneavenue is toexamine cross-linguistic variation amongsuperficiallytenseless languages.

• Today I’ll present four casestudies oflanguages which differ in theinterpretive possibilitie s theyallow for superficially tenseless clauses.

St’át’imcets (Salish) and Gitksan (Tsimshianic)

Blackfoot(Algonquian)Javanese (Austronesian)

Atayal (Austronesian)

6/81

Fourtypesofsuperficiallytenselesslanguage

St’át’imcets andGitksan:Superficially tenseless sentences (STSs)arestrictly non-future forallpredicate types andviewpoint aspects.

Blackfoot:STSsarestrictly non-future forallpredicate types, and inadditionperfective eventives areonly interpreted aspast.

Javanese:STSscanbepast, present or future, but there is optional overtpast tense.

Atayal:STSs in theactorvoicearestrictly non-future, butSTSs innon-actor voicescanbepast, present or future. (And, there’s optional overtpast tense.)

2

7/81

Theoverallproposal

• Thecross-linguistic variation we see isbetter captured byassuming that allthese four types oflanguage are tensed.

8/81

Thewarm-up: St’át’imcets andGitksan

Matthewson 2006.

Jóhannsdóttir &Matthewson 2007.

Matthewson 2013.Rullmann &Matthewson toappear.

9/81

St’át’imcetsandGitksan

St’át’imcets

• /šƛ’æƛ’yəmxəc /

• Also known asLillooet• Salish family, Northern Interior branch

• British Columbia, Canada

• Endangered

Gitksan

• Tsimshianic family• Dialectcontinuum

• British Columbia, Canada• Endangered

10/81http://www.bced.gov.bc.ca/abed/images/map2.jpg

11/81

St’át’imcetsandGitksan:STSsarenon-future

• Neither past norpresent tense is overtlymarked.STSscanbe interpreted aspast orpresent, with allpredicate types andviewpoint aspects.

Perfective eventive:

(1) Bax=t Yoko.run=DM Yoko‘Yoko ran’ / ‘Yoko is running.’ (Gitksan)

Stative:

(2) Siipxw=t James (k'yoots).sick=DM James (yesterday)‘James was sick (yesterday)’ / ‘James is sick.’ (Gitksan)

Progressive:

(3) Yukw=hl ga-gol-diit.PROG=CN DUR-run-3.II‘They were/are running.’ (Gitksan) 12/81

St’át’imcetsandGitksan:Futuretimereference

• Future timereference isobligatorily marked, inGitksan bydim.

(4) *(Dim) limx=t James t'aahlakw.*(PROSP) sing=DM James tomorrow James will sing tomorrow.ʼ (Gitksan)

(5) *(Dim) siipxw=t James t'aahlakw.*(PROSP) sick=DM James tomorrow‘James will besick tomorrow.’ (Gitksan)

(6) Yukw=hl *(dim) ga-gol-diit t'aahlakw.PROG=CN *(PROSP) DUR-run-3.II tomorrow‘They will be running tomorrow.’ (Gitksan)

3

13/81

St’át’imcetsandGitksan:Futuretimereference

• ‘Past future’ readings exist.

(7) Gilbil=hl ganuutxw=hl hli=daa=t mahl-i=s Diana dim wil yee=ttwo=CN week=CN PRT=SPT=3.I tell-TR=PN Diana PROSP COMP go=3.I

goo=hl Winnipeg ji hlaa (am) k'i'y=hl ganuutxw.LOC=CN Winnipeg IRR INCEP (only) one-CN week

‘Diana said twoweeks ago that she would go toWinnipeg afterone week.’

14/81

St’át’imcetsandGitksan:Tensedanalysis

• All finite clauses contain aphonologically null, non-future tense.

• Pronominal analysis forconcreteness. Tense isdecomposed into an indexedtemporal variable andapresuppositiona l feature (Cable2013,Mucha 2015,Bochnak2016).

(8) T’

Ti NON-FUTURE

• Tidenotes thecontextually provided RT.NON-FUTURE presupposes thatRTdoesnot follow UT,andotherwise denotes the identity function.

(9) ⟦ NON-FUTURE ⟧g,t0,w0 =λt :t≤ t0 . t

• Dim isaprospective aspect. Itco-occurswith tense, just likeWOLL inEnglish.

(10) ⟦ PROSP ⟧g,t0,w0 =λP<i,s t> λtλw .∃t [t<t &P(t )(w)]

15/81

Potentialtenselessanalysis1:Bohnemeyer(2009)

(11) Modal Commitment Constraint: The realization ofevents in the (relative orabsolute) future cannotbeasserted, denied, questioned, orpresupposed asfact. Assertions, questions, and presuppositions regarding the futurerealization ofevents… require specification ofamodal attitude … .

• InYucatec, theperfective conveysevent realization and is notcompatiblewith future interpretation inclauses which assert/question /pr esuppo se.Progressives andstatives allow future RTs:

(12) Táan in=mèet-ik le=nah=o’.PROG A1SG=do:APP-INC(B3SG) DET=house=D2‘I am/was/will bebuilding thehouse.’ (Bohnemeyer2009)

(13)Túumben le=nah=o’. new(B3SG) DET1=house=D2‘The house is/was/will benew.’ (Bohnemeyer2009)

16/81

DifferencesbetweenSt’át’imcets/GitksanandYucatec

• InSt’át’imcets andGitksan, evenstative and imperfective STSsare strictlynon-future in theabsence ofovertprospective.

• InSt’át’imcets andGitksan, even in theperfective, events canbe in thepresent without having tobe ‘blow byblow reports’, unlike in Yucatec(Bohnemeyer 2009:25).

17/81

Potentialtenselessanalysis2:Smithetal.(2005)

Pragmatic default principles:

i. DeicticPrinciple: Situations are located with respect toSpeechTime.ii. Bounded EventConstraint: Bounded events arenot located in thePresent.

iii. Simplicity Principle ofInterpretation: Choose the interpretation that requiresthe least information added or inferred. (Smithetal. 2005,Smith2007)

• Together, thesepredict:

• Unbounded eventualities are interpreted aspresent.• Bounded events (telicand/or perfective ones) are interpreted aspast.

Adifference betweenSt’át’imcets/Gitksan and Navajo:

• InSt’át’imcets andGitksan, perfective events arenot restricted topast tense,so theBounded EventConstraint must be removed.

18/81

Summary:Tenselesslanguagetype1

• STSsarestrictly non-future forallpredicate types andviewpoint aspects;overtmarking for futurity.

• Tensed analysis involves covertnon-future tense.

• Available tenseless analyses need tobeatleast tweaked if theyare tocapture the facts.

4

19/81

Controversy beginstoloom: Blackfoot

Reis Silva&Matthewson 2007.Ritter &Wiltschko 2014.

20/81

Blackfoot

• Algonquian family• Alberta, Canada andMontana, USA

• Endangered

21/81

Blackfoottemporalbasics

• Noovertpresent orpast tense morphology.

• Obligatory overtmarking for future time reference (áak- oráyaak-;ReisSilva2009).

(14) Anna Mai’stoo isttso’kiniann-wa Mai’stoo-wa isttso’kini-waDEM-3 Raven-3 hungry.VAI-3‘Mai’stoo is/was hungry.’ ≠‘Mai’stoo will behungry.’ (Louie2014:7)

(15) Anna Mai’stoo áíhpiyiann-wa Mai’stoo-wa á-ihpiyi-waDEM-3 Raven-3 IPFV-dance.VAI-3‘Mai’stoo is/was dancing.’ ≠ ‘Mai’stoo will dance.’ (Louie2014:8)

22/81

Grammaticalaspect

• Imperfective isovertly markedonallpredicate types (Dunham 2007).

• Predicates unmarked for imperfective areperfective.

Eventives:

(16) nítsspiyi (17) nitáihpiyinit-ihpiyi nit-a-ihpiyi1SG-dance 1SG-IPFV-dance‘I danced.’ ‘I amdancing.’/ ‘I dance.’ (habitual)

Statives:

(18) nitsikooksina (19) nitsikaoksin kam’so’ohkooyiinik init-ik-oksina nit-ik-a-oksin kamm-sa-ohk-ooy-yiniki1SG-INT-mean 1SG-INT-IPFV-mean if-NEG-all-eat-1SG.SBJ‘I ammean.’ ‘I ammean if I don’t getanything toeat.’

23/81

Arestrictiononpresenttenseinterpretations

• STSsallow either past orpresent interpretations only for statives andimperfectives.

• Perfective eventive STSsareobligatorily interpreted with pastRTs.

24/81

Acontextwhichforcesapresent-tenseinterpretation

Your friend calls youon thephoneandasks you tomeetwith her right now.Yourespond bysaying ‘I can’t meetwith youright now because …’

(20) nitsíksttsokini (21) nitáihkiitanit-ik-sttsokini nit-a-ihkiita1SG-INT-hungry 1SG-IPFV-cook‘I amreally hungry.’ ‘I amcooking.’

(22) nitaoksstoopa omi sinakiatsisnit-a-oksstoo-’p-wa om-yi sinaaki-a’tsis1SG-IPFV-read-LOC>0-3 that-0 write-tool‘I amreading that book.’

(23) # nitsikksstoopa omi sinakiatsisnit-ii-okstoo-’p-wa om-yi sinaaki-a’tsis1SG-IC-read-LOC>0-3 that-0 write-tool‘I read thatbook.’

• Sameresult with ihkiita ‘cook’, ihpiyi ‘dance’, iikooisska ‘build ahouse’,aasáini ‘cry’, iiyokaa ‘sleep’, … .

5

25/81

Incompatibilityofperfectiveswithpresenttemporaladverbials

(24) nitao’taksinaaki annohknit-a-o’tak-sinaaki annohk1-IPFV-round-draw now‘I’mdrawing acircle rightnow.’ (Dunham2007)

(25)* nitsi’taksinaaki annohknit-i-o’tak-sinaaki annohk1--round-draw now (Dunham2007)

(26)* annohk kitána aasáí’niannohk k-itána aasáí’ninow 2POSS-daughter cry‘Right now yourdaughter cried.’

Consultant’s comment: “Itis too past touseannohk.”

26/81

Blackfoot:Theargumentfortense

• The temporal restriction isparallel toa restriction inEnglish: eventivepredicates mustbe in the imperfective if the reference timecoincides withtheutterance time:

(27) I can’t meetwith you right nowbecause …

a. I ’mhungry.b. I’mcooking/building ahouse.c.# I cook/build ahouse.

Tonhauser (2015):

‘A tensed analysis ofatenseless language isempirically motivated if thelanguage exhibits temporal reference restrictions comparable to thoseexhibited bysome tensed language.’

27/81

Blackfoot:Tensedanalysis

• Null pastand null present.

(28) ⟦ PAST ⟧g,t0,w0 = λt :t<t0 . t⟦ PRESENT⟧g,t0,w0 = λt :t=t0 . t

• Eventive predicates require the imperfective in thepresent tense inBlackfootfor thesamereason theydo inEnglish (Bennett &Partee 1978):

• Apresent perfectivewould require the event to fit inside themomentofutterance (Klein 1994), butevents cannot fit inside moments.

• Statespossess thesub-interval property.Theycanhold atmoments,therefore canbe in thepresent perfective.

(29) ⟦PRES PFV Iread thatbook⟧g,t0,w0 =⟦PRES PFV nitsikksstoopa omisinakiatsis⟧g,t0,w0=∃e [read.that.book(e,w0) &Agent(I,e,w0) &τ(e)⊆ t0]

# ‘There is anevent ofmy reading thatbook whose run-time is included withinthemoment ofutterance.’

28/81

Potentialtenselessanalysis1:Bohnemeyer(2009)

• Modal Commitment Constraint plus theclaimthat theperfective conveysevent realization.

à Correctly derives the restriction ofperfectiveeventives topast.

Difference between Blackfoot andYucatec:

• InBlackfoot, evenstative and imperfective STSsare strictly non-future in theabsence ofáak or áyaak.

29/81

Potentialtenselessanalysis2:Smithetal.(2005)

i. DeicticPrinciple: Situations are located with respect toSpeechTime.

ii. Bounded EventConstraint: Bounded events arenot located in thePresent.iii. Simplicity Principle ofInterpretation: Choose the interpretation that requires

the least information added or inferred. (Smithetal. 2005,Smith2007)

à Correctly derives the restriction ofperfectiveeventives topast.

Issues:

• Wewould need toassume that stative predicates without imperfectivemarking arenot in theperfective. What is their viewpoint aspect, then?

• Theprinciples aresupposed tobepragmatic defaults, but theBounded EventConstraint isnon-cancellable. This seems tobecaptured better semantically(e.g.,byapresent tense morpheme) than byapragmaticdefault principle.

30/81

Potential‘tenseless’analysis3:Lin(2006)

• Lin (2006) forChinese:

(30) ⟦ Perfective aspect⟧ = λP<i,t> λtTop λt0∃t [t⊆ tTop ∧ P(t)∧ tTop< t0

‘This relation actually incorporates the notion of semantic tense into thesemantics ofaspect’ (Lin2006).

Issues:

• Instead ofhaving null present andnull past, would:i. havenull perfective which incorporates past semantics

ii. require us toassume thatstatives without imperfective aren’t perfective

iii. require a supplementary pragmaticprinciple to rule out futureinterpretations for statives and imperfectives.

6

31/81

Potentialtenselessanalysis4:Ritter&Wiltschko(2014)

• Thehead ofInfl inBlackfoot marks not temporal (non-)coincidence (=tense), but participant (non-)coincidence.

• Participant (non-)coincidence is markedby ‘order’ affixes: -hp for1st/2nd

person, Ø for3rd.

Problem:

• Fails tocapture the factthatperfective eventive sentences only havepasttense interpretations.

32/81

Potentialtenselessanalysis4:Ritter&Wiltschko(2014)

• Ritter &Wiltschko’s analysis isdesigned tocapture thegeneralization that ‘aclause that lacksovertm[orphological]-marking for tense is compatible witheither apresent orapast time interpretation’ (2014:1332).

• Their evidence for this generalization isas follows:

(31) Oma píítaawa áípaawaniwa.om-wa píítaa-wa a-ipaawanI-waDEM-PROX eagle-PROX IPFV-fly.AI-PROX‘That eagle is/was flying up.’ (Ritter&Wiltschko2014)

• (31) ishalf ofaminimal pair givenbyReis Silva&Matthewson (2007); theother half shows thataperfective version of(31)cannot haveapresent-tense interpretation.

• SinceRitter &Wiltschko donot address the temporal restriction onperfective eventives, their analysis does notcapture the factsabouttemporal interpretation inBlackfoot.

33/81

Summary:Tenselesslanguagetype2

• STSsarestrictly non-future forallpredicate types; overtmarking for futurity.

• Inaddition, perfective eventives areonly interpreted aspast.

• Tensed analysis involves covertpastvs.present tense.

• Available tenseless analyses need tobeatleast tweaked if theyare tocapture the facts.

34/81

Radical tenselessness that isn’t: Javanese

Chen, Matthewson, Rullmann &Vander Klok2017.

35/81

Javanese

• Austronesian (Malayo-Polynesian).

• Central and Eastern Java.• > 90million speakers, mostbilingual with Indonesian.

• Speech levels: ngoko ‘Low’, madya ‘Mid’, krama ‘High’. • Dialectgroups: West, Central, East (Hatley 1984).

NB:Everything in this section applies toAtayal too, but I ’msaving Atayal tomakeadifferent point in thenext section. SeeChen etal. (2017).

36/81

Javanese:temporalbasics

• Appears tobe radically tenseless: STSscanbe interpreted aspast, present orfuture.

(32) A: Wingi / saiki / sesok ewoh opo?yesterday /now /tomorrow busy what‘Yesterday what [were you]doing?’‘Now what [areyou]doing?’‘Tomorrow what [will yoube]doing?’

B: aku marut kelopo1SG AV.grate coconut‘I shavedcoconut’ / ‘I was shaving coconut.’‘I amshaving coconut.’‘I will beshaving coconut.’ (VanderKlok&Matthewson2015)

• ‘I suggest that Javanese isa tenseless language’ (Vander Klok 2012).

• But, weargue that Javanesepossesses anoptional past tense marker.

7

37/81

Javanesetau

• Tau is an auxiliary which hasadominant experiential reading. Dahl (1985)classifies tau asan‘experiential’ aspect.

• Experiential aspects are ‘closely related to’ perfects, according toDahl.

(33) A: Sampean tau m-(p)enek gunung Merapi toh?2SG TAU AV-climb mountain Merapi FOC‘Haveyoueverclimbed Mount Merapi?’

B: Iyo, aku tau gelek m-(p)enek gunung iki.yes 1SG TAU often AV-climb mountain DEM‘Yes ,I haveoften cl imbe dt hatm oun tain. ’

38/81

Dominantexperientialreadingsoftau

(34) Q: Opo awakmu weroh dulur-ku?Q 2SG know sibling-my‘Doyouknow mybrother?’

A: Yo, aku tau ketemu dulur-mu, dadi aku weroh dulur-mu.

yes, 1SG E.PAST meet sibling-your become 1SG know sibling-your

‘Yes, I ’vemethimso I knowhim.’ (contextfromDahl1985:#37)

• Inspite oftau’s dominant experiential reading, we argue that it is not anaspect, but is anexistential past tense.

• Tau is arelative tense, and this isdistinguishable fromaperfect aspect(supportingBohnemeyer2014).

39/81

Tau isnotanaspect

40/81

Previewoffindings

Javanese tau

(English) presentperfect

Experiential reading ✓

Resultative reading ✗

Universal reading ✗

Current relevance ✗

Lifetimeeffects ✗

Adverbial restrictions ✗

Perfective Narrative progression ✗

Aspect ingeneral Unrestricted RT ✗

41/81

Noresultative reading

• Result state reading ofEnglish perfect (Mittwoch2008,amongothers):

(35) I have lost mywatch (#but I found itagain).

• Taucanbeused with change-of-state verbs without any implication thatthe result statestill holds. Tau actually implies that the result state nolonger holds.

(36) Context: Nowhe isnotatWisata Bahari Lamongan (WBL).

Bapak-mu (wes) tau melbu nok WBL mbiyen. father-your already E.PAST enter at WBL before‘You rfat her ente red in toWBLinth ep ast.’

42/81

Nouniversalperfect

• Areading of theEnglish present perfect where theeventuality holdsthroughout an interval starting atsome point in thepast up to thepresent(e.g.,McCawley 1971,Comrie 1976,Iatridouetal. 2001,Portner2003).

(37) Ihavebeen sick since1990. (Iatridouetal. 2001:155)

• Tau does notallow this interpretation:

(38) Context: Youmoved toJember fromPaciran in 2014&youstill live therenow.

# Aku tau manggon nek Jember sampai 2014. 1SG E.PAST live in Jember since 2014Intended for ‘I have lived in Jember since 2014.’

8

43/81

Nocurrentrelevancerequirement

• TheEnglish present perfect indicates current relevance (e.g.,Portner 2003,a.o.),but tau is infelicitous incontexts highlighting thecurrent relevanceofthepredicate.

(39) Context: Your friendasks if youwant toeatatBuMaula’s. You finishedeating 10minutes ago.Yousay:

Sepura-ne, aku {# tau /✓wes } mangan. sorry-DEF 1SG E.PAST already AV.eat‘Sorry, I ’veeaten.’

44/81

Nolifetimeeffects

• With theEnglish present perfect, predicates with subjects thatarenolonger alive aregenerally unacceptable, but tau is felicitous.

(40)??Gutenburg has discovered theartofprinting.(McCoard1978,citingDietrich1955)

(41) Columbus tau nemok-no Amerika. Columbus E.PAST AV.find-APPL America‘Columbus found America.’

45/81

Nodefiniteadverbialeffects

• Tau is compatible with definite past-time adverbials, incontrast to theEnglish present perfect (Klein 1992).

(42)* Chris has leftNewYorkyesterday. (Klein1992)

(43) Aku tau mangan rajungan wingi wingi-nan-e. 1SG E.PAST AV.eat crab yesterday yesterday-NMLZ-DEF‘I atecrab2daysago.’

(44) Adik-ku tau lungo neng Indonesia september 2015. sibling-my E.PAST go to Indonesia September 2015‘Myyounger sibling went to Indonesia inSeptember 2015.’

46/81

Tauisalsonotaperfectiveaspect

• Perfective aspect is typically used fornarrative progression:

(45) Marywoke up.Sheatebreakfast andwent to thebeach.

• Taudoes notallow narrative progression:

(46)Context:Youaredescribing what happened yesterday.

Siti melbu kantor. De’e (# tau) ngopi. Siti AV.enter office 3SG E.PAST AV.coffee‘Siti cameto theoffice.Shedrank coffee.’

47/81

Taudoesn’tallowfreeReferenceTimes,likeaspectsshould

• Ina (neo-)Reichenbachian/Kle in ian(Pa ul ian!) system, aspects donotdirectly locateRTs on the timeline; that isdone by tense.

• Aspects should beable toco-occurwith different tenses, allowing differentRTs.

(47) a. I hadeaten crab. Past perfectb. I haveeatencrab. Present perfectc. I will haveeaten crab. Future Perfect

• Tau is notcompatible with present or future reference times, asdemonstrated by restrictions onadverbs:

(48) Aku tau mangan rajungan wingi / # saiki / # sesok.1SG E.PAST AV.eat crab yesterday / now / tomorrow‘Iatecrabyesterday.’ /≠ ‘I haveeatencrabnow.’/ ≠ ‘I will haveeatencrabtomorrow.’

48/81

Summaryoffindings

Javanese tau

(English) presentperfect

Experiential reading ✓

Resultative reading ✗

Universal reading ✗

Current relevance ✗

Lifetimeeffects ✗

Adverbial restrictions ✗

Perfective Narrative progression ✗

Aspect ingeneral Unrestricted RT ✗

9

49/81

Tau isanexistential pasttense

50/81

Pronominalvs.existentialpasttenses

• Thepronominal analysis I ’veassumed so far (Partee 1973, Heim 1994,Kratzer 1998 a.o.)contrasts with anexistential quantifier analysis (Ogihara1996, vonStechow 2009, a.o.).*

Existential past tense:

(49) ⟦ PAST ⟧g,t0 =λP .∃t’ [t’< t0 &P(t’)]

• Empirical evidence todistinguish these two types ofpast tense is subtle.

• TheEnglish pasthas beenanalyzed bothways, and alsoasbeing ambiguousbetwe en the two(vonStechow&Grønn2013a,b,Grønn&Stechow2016,a.o.).

* Theview that tenses areoperators which relate two time intervals given inthesyntax (Zagona 1990, Stowell 1993, 2005, Demirdache &Uribe-Etxeberria 1997, 2007, 2014, …) is in relevant respects avariant of thepronominal approach.

51/81

Distinguishingthetwotypesoftense

Thepronominal analysis predicts that past tenses:

• Don’t allow scopal interactions with negation

• Allow anaphoric and deicticuses

• Are infelicitous without acontextual reference time (setting aside bounduses).

Theexistential analysis predicts thatpast tenses:

• Allow scopal interactions with negation• Lackanaphoric ordeictic uses

• Are felicitous discourse-initially

52/81

Testingthepredictions1:Scopeinteractions

(50) Context: Drivingon thehighway after leaving thehouse.

I didn’t turnoff thestove. (Partee 1973:602)

i. ∃t [t<now&¬ [I turnoff thestove att ]] existentialii. ¬∃t [ t<now& Iturn off thestoveat t ] iii. ¬ [I turn offthestove at t],where t<now pronominal

Partee’s argument:

• (50i)would mean:There is sometime in thepast atwhich I didn’t turn offthestove (tooweak).

• (50ii) would mean:There isno time in thepast atwhich I turned off thestove (toostrong).

• (50iii) means: atsome contextually salient past interval, I didn’t turn offthestove (just right).

53/81

Testingthepredictions1:Scopeinteractions

• Tau has the scopal readings thatEnglish lacks!

• When negation scopes over tau, we getthe¬>∃ (‘never’) reading:

(51) Wong londo gak tau mangan sego.person foreigner NEG E.PAST AV.eat rice‘Foreigners havenever eaten rice.’¬∃t [t<UT& [foreigners eat riceatt]]

• When tau scopes overnegation, we gettheexpected∃> ¬ reading:

(52) Context: Wanan eats rice everyday.Butmaybehehasn’t eaten riceonce ortwice.Pak Wanan tau gak mangan sego.Mr. Wanan E.PAST NEG AV.eat rice‘PakWanan has noteaten ricebefore.’ ∃t [t<UT& ¬ [Wanan eat riceat t]]

54/81

Testingthepredictions2:Noanaphoricuses

• Thenarrative progression dataprovide evidenceagainst anaphoricinterpretations oftau.

• Insimple linear narratives, apasteventive sentence typically “updates” theRT toatime just after theRTwhich isprovided bypreceding discourse(Partee 1984,Hinrichs1986,Kamp&Reyle 1993,a.o.).

• Tau is rejected innarrative progression contexts:

(53)Context:Youaredescribing what happened yesterday.Siti melbu kantor. De’e (# tau) ngopi. Siti AV.enter office 3SG E.PAST AV.coffee‘Siti cameto theoffice.Shedrank coffee.’

10

55/81

Testingthepredictions3:Nodeicticuses

Context: Drivingon thehighway after leaving thehouse, you realize thatyoudidn’t turn offthestove.

(54) Aku kok rung (# tau) mate-ni kompor yo!1SG PRT not.yet E.PAST AV.die-APPL stove yes‘I didn’t turn off thestove!’

è Onalldiagnostics, taubehaves likeanexistential rather thanapronominaltense.

56/81

‘Experiential’readingsasexistential

• Wesaw at thebeginning that tauhasa salient experiential reading.

• Experiential readings areexistential readings.

• Tau isn’t restricted topure experiential readings; the times overwhich itquantifies canbenarrowed down bypast timeadverbs.

(55) Adik-ku tau lungo neng Indonesia september 2015. sibling-my E.PAST go to Indonesia September 2015‘Myyounger sibling went to Indonesia inSeptember 2015.’

• Thedomain of theexistential quantification canbe restricted.

57/81

Ouranalysis

(56) ⟦ 𝑡𝑎𝑢 ⟧ =λC<i,s t> λP<i,s t> λt λw .∃t’ [t’< t&C(t’)(w) &P(t’)(w)]

• The time intervals overwhich tau quantifies aredomain-restricted (via theCvariable) (as invonStechow 2009).

• Its targument is saturated by themost contextually salient time, which inmatrixclauses is theutterance time.

58/81

Thisanalysiscapturesallthesefacts:

Javanese tau

English presentperfect

Experiential reading ✓

Resultative reading ✗

Universal reading ✗

Current relevance ✗

Lifetimeeffects ✗

Adverbial restrictions ✗

Perfective Narrative progression ✗

Aspect ingeneral Unrestricted RT ✗

Existential vs. pronominal past

Scope interactions ∃

Anaphoric uses ∃

Deicticuses ∃

Optional ✓

59/81

Afurtherprediction:Cessationinferences

• Past tenses arewell-known togive rise tocessation inferences with statives(Musan1997,Magri2009,Thomas2012,Altshuler &Schwarzschild2013,Cable2016,a.o.).

(57) A: How is Scottydoing?B: Hewas anxious. (Implies Scotty is no longer anxious)

(Altshuler &Schwarzschild2013)

• Taugives rise tocessation inferences:

(58) Context: Mrs.Siti is nowslim.

Bu Siti tau lemu.Mrs. Siti E.PAST fat‘Mr s.Sit i wa sfat.’ (Speakers all comment thatMrs.Siti isno longer fatat theutterancetime.)

60/81

Cessationinferences

• Roughly following Altshuler &Schwarzschild, Cable:with states, presententails past, sopast implicates not-present.

• Even though Javanese doesn’t haveapresent tense, wecanget thecessation inference to runoff implicature ifweanalyze the language ashaving apronominal covert tense which by default is interpreted asreferring toUT.

• Taucompetes with this pronominal tense, giving rise tocessationimplicatures.

11

61/81

Tau isarelativepasttense

62/81

Relativetensevs.perfectaspect

• Comrie (1985):Relative tenseneed not locate asituation relative toUT.Ithasbeen used to refer tonon-finite haveand ‘past-in-the-past’ have (seealsoStowell, thisworkshop!).

(59) Having left earlier, John took the bus. (60) John had already leftat10pm.

Relative tense separate fromperfectaspect?

• No forKlein (1994)• No/yes forArregi &Klecha (2015),Klecha (2016)• Yes forBohnemeyer (2014)

• Bohnemeyer: True relative tense relates RT toanevaluation time. (i.e.,there are four relevant time intervals inaneo-Reichenbachian system).

63/81

Evidencethattauisarelativetense

• When embedded underattitudes/reports , tau cannot receivesimultaneousbutonly back-shifted interpretations:

(61) Pak Agus ngomong deke tau nesu.Mr. Agus AV.say 3SG E.PAST angry‘Mr.Agus said thathewasangry.’

# Simultaneous context:Agus was scheduled tomeetwith Eko at10amyesterday.But at1pm, Eko was still not there.Agus called mebecause hewas angry.Then, I told myneighbour (61).

✓ Backshifted context:Agus wasangry last week.Hecalled meyesterdayafternoon to tell methathehadbeen angry.

64/81

Pastrelativetomatrixfuture

(62) Context: Youencourage Siti toworkonher thesis this afternoon, eventhough it is implausible that shecan write thewhole thesis. “After all,Mother will know youhaveworked”, yousay.

--------UT--------work--------know--------

Ibuk-mu ape ngerti awakmu tau nggarap skripsi-mu.mother-your FUT know 2SG E.PAST AV.make thesis-your‘Your mother will know youworked onyour thesis.’

è Tau markspast relative tosome evaluation time, rather than toUT.Relative past isdistinct fromperfectaspect.

65/81

Summary:Tenselesslanguagetype3

• STSscanbepast, present or future and the language seems at first glance tobe radically tenseless.

• Closer examination reveals that there is optional overtpast tense.

• Tensed analysis hasoptional existential relative tense, plus covertpronominal tensewhich bydefault picksout UT.

66/81

Potentialtenselessanalysis?

• Tenseless analysis seems impossible, because tau is atense.

• Theonly option would be tostipulate that tenses can’t beoptional, andtherefore that tau, although itpossesses exactly thesemantics ofapasttense, must becalled something different.

• Lin (2012): ‘A useful criterion …is that tense morphemes are integrated intothegrammarof the language, typically morphologically bound, andareobligatory, even though theyarenot necessary for interpretation.’

? Why?

12

67/81

Theclincher: Atayal

Chen inprep.

68/81

Atayal

• Austronesian.

• Northern and northeast Taiwan.• Endangered.

69/81

Atayal:temporalbasics

• Austronesian voicesystem; relevant here isactorvoicevs.non-actor voices.

Actor voice:

• All actorvoiceSTSsarestrictly non-future. Future interpretations requireovertmarking (byprefixing p-,orwith theauxiliary musa’).

(63) m-’uy=saku’ la. AV-tired=1SG.ABS PRT‘I was tired.’ / ‘I amtired.’ /=‘I will be tired.’

(64)* m-t-zyuwaw=saku’ kira’AV-ATR-work=1S.ABS today.laterIntended for ‘I will work later.’

• Eventives don’t allow present episodic interpretations (likeBlackfoot).

(65) m-nbuw hiya’. av-drink 3s.n‘Heused todrink.’ / ‘Hedrinks.’ [habitual] /= ‘Hewill drink.’ 70/81

Atayal temporalbasics

Non-actor voice:

• Radical tenselessness. Aspectually unmarked non-actor voicepredicates canbe interpreted as (past, present or) future without overt temporal marking.

(66) Context: Children areplaying balls near thewindows. Youask themto leave.bkawn=mamu’ tubung lki!break.PAT.V=1P.ERG window PRT‘You will break the windows!’

• The future interpretation isnot restricted toplanning contexts oraparticular modality.

71/81

Atayal:Tensedanalysis(Cheninprep.)

• Covert non-future tense.

• Prospective aspect, spelled outeither asp-/musa’ (actorvoice)orØ (non-actorvoice).

• (Seealso Bochnak2016 forapartially similar approach topartially similarfacts inWasho.)

Potentialtenselessanalysis?

? Can’t think ofone.

72/81

Conclusions

13

73/81

Fourtypesofsuperficiallytenselesslanguage

St’át’imcets andGitksan:Superficially tenseless sentences (STSs)arestrictly non-future forallpredicate types.

Blackfoot:STSsarestrictly non-future forallpredicate types, and inadditionperfective eventives areonly interpreted aspast.

Javanese:STSscanbepast, present or future, but there is optional overtpast tense.

Atayal:STSs in theactorvoicearestrictly non-future, butSTSs innon-actor voicescanbepast, present or future. (Plus there’s anoptional overtpast tense.)

• Ihaveargued foratensed analysis ofeach,based ondifferent language-internal evidence ineach.

74/81

Fourtypesofsuperficiallytenselesslanguage(simplified)

STSsnon-future STSs free

PFVeventives past BlackfootAtayal actor

Yucatec

PFVeventives free St’át’imcetsGitksan

JavaneseAtayal non-actor

75/81

Cross-linguisticvariationspeaksagainstpragmaticprinciples

• Example:The restriction against present-tense perfectiveeventives inBlackfoot andother languages.

• St’át’imcets lacks this restriction:

(67) Context:Your friend calls youupandwants you tomeether rightnow.Ao kw=en ka-nás-a áku7 snúwa …NEG DET=1SG.POSS CIRC-go-CIRC DEIC you…‘I can’t cometoyourplace…’

a. máys-en=lhkan ti=n-q’íl’q=a.fix-DIR=1SG.SBJ DET=1SG.POSS=chair=EXIS‘I ’m fixing mychair.’ PERFECTIVE

b. k’wezús-em=lhkan.work-MID=1SG.SBJ‘I ’mworking.’ PERFECTIVE

76/81

Cross-linguisticvariationspeaksagainstpragmaticprinciples

• TheSt’át’imcets factsarepredicted bythenon-future tense analysis: sincethere isno instantaneous present tense in the language, eventives can fitinside atime interval which includes UT.

• It’snot impossible that languages varyin thegeneral pragmatic principlestheyadopt, but this cross-linguistic variation at least suggests thatwecan’tassume restrictions of theBlackfoot type follow for free.

• And remember Atayal, where the interpretation ofSTSsvaries within onelanguage, according to thevoiceparadigm.

77/81

Howtensescanvary

• Overtness

• Pronominal vs.existential semantics• Optionality

• Theprecise presupposition s theyplace on theRT (past, non-future, etc.)

Doweneedtoworryaboutanyofthis?

• Covert morphology isprevalent in analyses of theworld’s languages.

• Functional elements canbeoptional (e.g.,determiners).

• Different levels ofprecision/underspecificat ion forpresuppositionalfeatures also exist, e.g.with pronouns.

• Ritter &Wiltschko’s Participant-based analysis ofBlackfoot involves covertmorphology (3rd)andsemantic underspecification (1st/2nd neutralized).

• Is thereaproblem with morphology that is bothcovertand underspecified?Why?What counts as toounderspecified, or specified enough?

78/81

Bohnemeyer’s challenge

• Bohnemeyer argues that some languages mark tense and others don’t, anddraws aparallel with the factthat some languages markdefiniteness andothers don’t, ornoun classes, orevidentiality.

‘The reason for this kindofcrosslinguistic variation in the functionalcategorysystem seems tobe that theexpression of functional categoriessuchas tense, viewpoint aspect, definiteness, gender, noun class, andevidentiality isnotnecessary for conveying the intended communicativecontent oflinguistic utterances. The relevant conceptual distinctions aremadewhether ornot theyareexpressed linguistically andspeakers canrelyonpragmatic means tocommunicate them where needed.’

• Iwouldn’t want tosay thatall languages havedeterminer, noun classorevidential systems even ifwe don’t seeovertevidence for it. Somelanguages do lackdeterminers, noun classes orevidentials.

• Thequestion then is: Is tensedifferent?

14

79/81

Istensedifferent?

• Ithink it is.

• Wecangetawaywith not specifying evidential values, butwecan’t getawaywithout havingRTs.

• The literature onsuperficially tenseless systems spends a lotof time derivingtemporal effects through other means.This is something thatwe just don’thave todo fordefiniteness orevidentiality.

• Assertions canbeneutral with respect todefiniteness orevidentiality inaway that theycan’tbe forRTs.

80/81

Recenttensedanalysesofradicallytenselesslanguages

Mucha (2015:70):

• ‘Formally, this is to saythat Hausapatterns with truly tensed languages inthat its structure contains anopen RTvariable whose interpretation dependsonacontextually defined assignment function.Thecrucial difference,however, is that the assignment ofavalue to this variable isnot restricted bythesemantics oftense morphemes, as it is in English or St’át’imcets.’

• ‘Why would I assume asyntactic RTvariable and thus abstain from makingthestronger claimthat tense isnot present in the structure ofHausaatall?On theone hand, conceptual considerations motivate thedecision toassumea timevariable in the syntax.…Moreover, there areempirical observationsthataremoreeasily accounted foron the assumption ofareference timevariable in the structure ofHausa…’ (SequenceofTense facts).

• Seealso Bochnak (2016) onWasho.

81/81

Thanksto

• St’át’imcets consultants the lateBeverley Frank, the lateGertrude Ned, LauraThevarge and the lateRose Agnes Whitley.

• Gitksan consultants Barbara Sennott, Vincent GogagandHectorHill.

• Javanese consultants Titis Subekti, Fina Aksanah, Bahrul Ulum, Nasrullah, LijatisHakim, Wuri Sayekti andAhlis Ahwan.

• Squliq Atayal consultants Heytay Payan, Buya’ Bawnay, and Maray Pasan.

• Co-authors Amélia Reis Silva, Sihwei Chen, Kristin Jóhannsdóttir, Hotze Rullmannand Jozina Vander Klok.

• Audiences atAFLA24, TripleA 4andLABG2017, andparticipants in theUBC TAP(Tense andAspect in thePacific) Lab.

• TheSocial SciencesandHumanities Research Council ofCanada.

• TheTenselessness Workshop for the invitation, and especially Maria Arche.