Inflec&onal Morphology and the Developmental Sciences

Farrell AckermanUniversity of California at San Diego

1

If you want to find new ideas, read old books:  not only old linguis7cs books, but old books in compara7ve animal behavior and ontogene7c development.Adaptedfromsomewhere.

The fact about science is that everyone who has a made a serious contribu7on to it is aware, or very strongly suspects, that the world is not only queerer than anyone has imagined, but queerer than anyone can imagine.  This is a most disturbing thought, and one flees from it by sta?ng the exact opposite.  J.S.HaldaneascitedR.G.Reid2007:4311

1.Reid,,G.B.2007.Biological Emergences: Evolu7on by Natural Experiment.  TheViennaSeriesinTheoreDcalBiology.MITPress.

2

Q:Whatdoyoudoifyoudon’twanttoflee?

General Task

Whatistherelevanceofthedevelopmentalsciencesforlanguageanalysis?

    What I am sugges7ng: 

     1. Developmental sciences (both psychological and biological) offer conceptual insights         about how to approach and explain objects within complex adap7ve systems.

     2.  Developmental sciences offer toolkits for analyzing such objects. 

     3.  When inflec7onal morphology is conceptualized as a complex adap7ve system, then           the relevance of the developmental sciences become evident.

     What I am not sugges7ng: 

     1. Language, in par7cular morphology, derives from biologically guided domain‐specific          species‐specific proper7es of humans.

     2. I don’t know whether this is so, but I do know that it is common in the developmental         sciences to recognize that looking for ins7ncts for X is oSen an obstacle for          understanding traits and behaviors (Y‐Z Kuo 1932; Schneirla 1966; Lehrman 1953, 1960, 1970; Go_lieb            1997; Amundson 2005; Blumberg 2009; Jablonka & Lamb 2005; Gilbert and Epel 2009, Hood et. al. 2010, among            others.)

3

Today’s goal

Focuson

1)theinstrucDvenessofhowdevelopmentalsciencesanalyzefamiliarphenomenaand“anomalous”phenomenawithoutanalyDcallyprivileginganyparDculartypeovertheother;

2)waysinwhichmorphologicalphenomenacanbeexplained/moDvatedbyviewingthemintermsoftheparDculargrammarsystemsinwhichtheyoccur;

3)waysinwhichtheorganizaDonofmorphologicalsystemscanbemeasuredandmodelledusingthetoolsstandardlyemployedfortheanalysisofcomplexobjects.

Standard tools and strategies of analysis in the developmental scienceWhat are some modern tools and insights for morphological analysis over words and paradigms?

1.RecogniDonoftheinstrucDvevalueofidenDfyingmorphologicalanomalies,i.e.,phenomenathatdon’tcomportwithexpectaDonandwhichraisequesDonsconcerningpossiblemorphologicalsystemsandtheirorganizaDon.

2.Appealtomechanismsofdevelopmentforexplainingwhycertainmorphologicalpa\ernsprobabilisDcallyarise,givenamulDplicityofinteracDngfactorsoverrealDme.

3.StaDsDcalandInformaDonTheoreDcmeasurement,i.e.,condiDonalentropymeasurestherelaDveinformaDoncontentofelementsinpa\ern‐basedsystems,idenDfyingtheimplicaDverelaDonsbetweenwordsandtherebythesystemicorganizaDontheyparDcipatein.

4.(Agent‐based)SimulaDonspermitmechanisDcmodelingofcomplexsystemsinordertotesthypothesizedmechanismsandpathsofdevelopment.(J.M.Epstein1999,

2006;S.Kirby2001;S.Kirby&J.Hurford2002,Wedel2004,2009,Oudeyer2006,amongothers)

A stone

                                                            Nothing is built on stone; all is built on sand.  But we must build as if the sand were stone. 

JorgeLuisBorges

A basic ques&on:       

What should count as stone for purposes of theore7cal analysis in morphology?

A stone:Wordsandparadigmsaretheprimaryobjectsofpa_ern‐theore7cmorphologicalanalysis.

LargelyarehabilitaDonofoldideaswithsomenewinsightsandtheuseofquanDtaDveandcomputaDonaltoolkits.(Paul1880;Robins1959;Ma\hews1965,1972,1974;MorpurgoDavies1978,1998;Wurzel1989;Anderson1992;Aronoff2004;Stump2001;Pirrelli2000;Gurevitch2000;Milinet.al.2009;Kro\et.al.2001;MoscosodelPradoMarDnet.al.2004;HayandBaayen2005;BlevinsandBlevins2009;Booij2010;amongothers

7 guiding hypotheses

H1:WordandParadigmMorphologyisusefullyviewedastypeofPa\ern‐TheoreDcMorphology.

H2:Theword(realizedbyenDDesparDcipaDnginagradientbetweenfullysyntheDcandfullyperiphrasDcexponence),insteadofthemorpheme,iscommonlythesmallestmeaningful(complex)structureinmorphology;

H3:Apa\ern‐theoreDcperspecDvetreatsmorphemes,wheretheyexist,assimplythemosttransparentpa\ernofform‐meaningmappingassociatedwithwordsconceivedas(complex)pa\erns(Blevins2010).

H4:Whentheelementsintheseformalpa\ernsarenotanalyzableasmeaningfulbits(morphemes),asfrequentlyoccurs,itistheconfigura7onal pa_ernwhichmadeupofsuchelementswherethemeaningofthewholereliablyresides,i.e.,wordsarerecombinant gestalts thatreusethesamepiecesinnewpa\ernsfornewpurposese.g.,Harris’Georgian“unwillingnessconstrucDon”)

7

7 guiding hypotheses

H5:WordsarefundamentalunitsforparadigmorganizaDonanditisparadigmaDccontrastbetweenwordsthatpermitsmorphologytobethestudyofcovariaDonbetweenwordshapepa\ernsandwordmeanings,withoutanecessaryappealtoclassicalmorphemes;

H6:ItistheimplicaDonalorganizaDonthatinheresinparadigmsbetweenwordsthatpermitsspeakerstoaccuratelygeneratenovelformsforknownlexemes;

8

7 guiding hypotheses

H7:Pa\ern‐theoreDcapproachestomorphology(WordandParadigmMorphologies)parDcipateinareconceptualizaDonofgrammaranalysiswhichviewsmorphologicalsystemsascomplexadapDvesystems.ThispermitstheuseofmanyoftheassumpDonsandtoolkitsthatguidetheongoingreconceptualizaDonofcomplexstructuresin

DevelopmentalComparaDvePsychologyandPsychobiology(D.Lehrman1953,1970;Y‐ZKuo,1960;G.Go\lieb1970,1992,1997;Karmiloff‐Smith1992;E.Thelen&L.BSmith2003;G.Michel&C.Moore1995;J.Elmanet.al.1997;J.SDles2008;;M.Bornstein&M.E.Lamb;2010;K.Hoodet.al.2010,

amongothers)

and

(Ecological)DevelopmentalEvoluDonaryBiology(LudwigvonBertalanffy1933/1962;PaulWeiss1939/196);SewallWright1968;S.Gilbertet.al.1996;LewonDn2001,S.B.Camazineet.al.2003;S.Oyamaet.al.2003;M.J.West‐Eberhard,2003;Amundson2005;S.Gilbert&D.Epel2009;A.Wallace2011;Bateson&Gluckman2011,amongothers.)

AdeeplydifferentconcepDonoftherelaDonshipbetweenperspecDvesonanalysisandrelevantmethodsofanalysisinthedevelopmentalsciencesandlanguagethanfoundinconvenDonalMainstreamGeneraDveresearchprogramse.g.,BiolinguisDcs.(Burraco&Longa2010;Bateson&Gluckman2011,amongothers.)

9

Organiza&on of the presenta&on

1.Compe7ng conceptualiza7ons of morphology (and grammar, more generally):    Morpheme‐Based versus Pa_ern‐Based approaches

Moro(Kordofanian)variablemorphotacDcs

2.The lessons from linguis7c anomalies:  extending the possible

Cwaya(Kordofanian)SUBJ/OBJreversal

3.Systemic explana7on: poten7a7ng the possible

TundraNenetsprenominalrelaDveclauses

4.Measuring and iden7fying the organiza7on of paradigm systems

LowcondiDonalentropyconjecture

5.ConcludingobservaDon

10

1. Compe&ng conceptualiza&ons of morphology 

Empiricalproblem:TherearesomemorphologicalsystemswhichdisplayvariablemorphotacDcsofmarkersinternaltowords.

GiventherecogniDonofthisphenomenonhowshouldweaccountforvariableaffixorder?(Luutonen1997;Muysken1988;J.Blevins2001;Bickelet.al.2006;Paster2005;Caballero2010;Stump2006;

Wa\ers2006;McFarland2007;Manova&Aronoff2010SpecialIssueofMorphology,amongothers)

11

Variable morphotac&cs on Thetogovela Moro (Kordofanian) Apprecia&on to Elyasir Julima and Ihklas Elahmer as our consultants

Basic morphotac7cs of Thetogovela Moro verb:

 {SM1ST&2ND‐}CM3RD.SM‐CLAUSE‐[OM‐ASP‐ROOT‐EXT‐ASP/MOOD] MACROSTEM‐OM‐OM‐OM.INST‐OM.LOC

Verbs with object markers (OM) display variable morphotac7cs for the OM:  `pull’

OM appearsaqertheverbsteminthePerfecDve,butbothbeforeandaqeritintheImperfecDve;3rdplural‐loisalwaysaqertheverbstem.

OM Perfective Imperfective3SG NON-HUMAN g-a-vəleð-ó g-a-vəléð-a1SG g-a-vəleð-í-ɲé g-a-ɲé-vəleð-a2SG g-a-vəleð-á-ŋá g-a-ŋá-vəleð-a3SG HUMAN g-a-vəleð-ó-ŋó g-a-ŋó-vəleð-a1 INC. DUAL g-a-vəleð-ə-nda g-á-ndə-vəleð-a1 EX. PL g-a-vəleð-álánda g-a-ɲə-vəleð-álánda1 INC. PL g-a-vəleð-ə-ndŕ g-á-ndə-vəleð-a-ŕ2PL g-a-vəleð-ə-nda g-á-ndə-vəleð-a3pl g-a-vəleð-ə-lo g-a-vəleð-a-lo

= Paradigmatic Gap

12

Compe&ng analy&c alterna&ves

Q:ShouldweregardthealternaDveordersas

1)derivablefromsomebasicorderor,

2)simplyasco‐equalconstrucDonalpa\ernsservingasexponentsfortonepa\ernsanddifferent/disDncDvemorphologicalpropertysets(RoseandJenks2011:underreview)?

AlternaDveanalyDcopDonsrepresentdifferentbetsandvisionsabouttheroleoflanguagevariaDoninunderstandingmorphology.

13

Two analy&c op&ons, stated baldly

H1:Mainstream Genera&ve Grammar:VariaDonrepresentsdeviaDonfromsomeselectedcanonicalorbasicform,itself(possibly)represenDngUNIVERSALGRAMMAR.

Q:So,whichorder,ifeither,isbasicinMoroandwhatarethegrammarprinciplesresponsibleforproducingthederivedorder(s)?

DesignedtobepredicDve,withunpredictedpa\ernsconstruedasprovidingnewinsightsintomorphologicalorganizaDon,ratherthanrousingLounsbury’s(1953)suspicionsaboutthetheoreDcalusefulnessofa“ficDveaggluDnaDveanalogue”.

14

Current morphemic approaches

“...wordsdon’texist,theyarethephlogistonoflinguisDcs.Everyonemustsaythatsentencesarebuiltoutofmorphemes...We...expectahighdegreeofisomorphism,ofthetypeexpressedbyBaker’s(1985)MirrorgeneralizaDon;althoughvariousaffixalproperDesmayleadtoreadjustmentrulesthatendupmaskingsyntacDcstructures;...Theextentofsuchreadjustmentrulesenrichesthemorphologicalcomponentofthegrammar,(Boeckx,Bare Syntax2008:63‐64)

AstonishingHypothesis:DeviaDonsfromexpectaDoncountasdiscoveriesthat“enrich”ourunderstandingofthemorphology.PresumablythegreaterthedeparturefromexpectaDons,themoreprofoundtheenrichment.Couldn’ttheneedforsuchenrichmentcountasevidenceagainstthebasicprogram?DoesnotexplaintheactualrarityofsystemaDcandwidespreadaggluDnaDonnor

morphotacDcvariaDon.

“Anomalies” teach us about the limits and nature of possible departures from the core, but they have little bearing on our beliefs about the essential nature of the core as morpheme-based. (language particular surface exponence is mere adornment).

Two analy&c op&ons, stated baldly

All(core)a\estedpa\ernsarepositedtofollowfromtree‐theoreDcarchitecturalassumpDonsaboutlanguagedesignguidedbyaninnatelanguagefaculty.

Eschewsredundancyasabasicpropertyofmorphologicalsystems.

Words and paradigms are epiphenomena

16

Two analy&c op&ons, stated baldly

H2:PaSern‐theore&c (Word and Paradigm):VariaDonssimplyrepresentalternaDverealizaDonsforform‐meaningmappings,ofgreaterandlessercomplexity,withnonehavingaparDcularlyprivilegedstatus.

Q:So,whatmoDvatestheparDcularMoroconstrucDonalpa\ernsandwhatdothesepa\erns,however,(un)familiartellusaboutnaturallanguagemorphology?

Isn’tdesignedtopredictspecificvariaDons,largelytodescribethem,promoDnganextremelyflexibleformalism,sufficienttodescribealla\estedpa\ernsandconstrainedtoreflectsystemicandconDngentproperDesofgrammarsaswellashuman‐specificcapaciDes,ratherthaninnatearchitecturallanguagestructures.

PermitsredundancyinrepresentaDonsandfrequency‐sensiDvestorageofexemplars.

         Words and paradigms are primary objects of theore7c                                                analysis  

17

Choosing H2

Basic General Strategy:Provide detailed descrip7ons of cross‐linguis7cally recurrent gramma7cal phenomena in all of their variety (without arbitrarily privileging any par7cular encoding).

    “Anomaly”doesn’texistinlanguage,rather,lurkingbehinditareanomalouspresupposiDonsandconvicDonthatobtaininlinguisDctheory.A.Kibrik2003:304

     Constant Large Ques&on: What are the bounds of variability and what constrains     it?

ThisalternaDveperspecDveturnsapparentlyunrulyrarityingrammarintoinstrucDveguidanceaboutthenatureofadequatelinguisDcarchitectures.

Recombina7on of individual elements and ensembles of elements found in independent construc7ons are systemic redeployments of old elements within new configura7ons constrained by con7ngent proper7es and shaped by human cogni7ve and perceptual capaci7es.

18

2.  The lessons from grammar anomalies: extending the possible

AnomalieshaveaninstrucDveroleinbiology

DespitetheircharacterizaDonaserrorsofnature,theanomalous,whenproperlyconsidered,forceustoconfrontandcorrectthoseerrorsinourthinkingthatoqenimpedescienDficinsightandprogress.”M.BlumbergFreaks of Nature:  What anomalies tell us about development and evolu7on.  OxfordUniversityPress2009:13

19

Guiding intui&on: Recombinant poten&al in biological forms

Platypus:Afurry,egg‐laying,duck‐billed,echolocaDng,venomous(whenmasculine)creature.

TwosillyquesDons:

1.Is the platypus more or less natural than the duck or the                                    beaver?                             2.  Is the platypus a departure from a canonical duck or                                  beaver, or is it the other way around?

While all of the basic platypus proper7es are familiar, their alignments produce novel, historically con7ngent, and ecologically viable pa_erns. 

The pa_ern is new, not the pieces (though, of course, the pieces themselves show varia7on).

“Anomalies”  extend our no7ons of what is possible, and hence, natural.

20

Guiding intui&on: Recombinant poten&al of grammar

...bothkangaroosandplatypuswererepresentaDveratherthanidiosyncraDcanomalies...‐H.Ritvo.The platypus and the mermaid and other figments of the classifying imagina7on. HarvardUniversityPress.1997:6

Similarlyconstruedasrepresenta&ve rather than anomalous paSerns,lessfamiliarmorphologicalconstrucDonsa\esttologicalpossibiliDesthatarenotgenerallyimaginedorpredicted.

CountsassuggesDveevidenceagainst“simplifying”uniformityassumpDons.

GrammaDcalplatypus:reuseofpiecesinnewconfiguraDonsfornewpurposes‐pronounincorporaDon,tone,grammaDcalfuncDonassignment,person‐hierarchy.

21

Recombinant parts and paSerns: Gramma&cal func&on marking within the verb (Ackerman & Rose in prepara&on)

Goal:  Contrast gramma7cal func7on marking in two related languages which exhibit resemblances and striking dissimilari7es.

Gramma7cal Func7on encoding within the word:  IncorporatedpronounsinMorohavededicatedformsandposiDonsfortheexpressionofSUBJandOBJ.

Gramma7cal Func7on ReversalconcernstheassociaDonofCMandPNMslotswitheitherSUBJorOBJfuncDons,dependingonco‐occurringfeaturesetsandtonewithintheword.

22

Gramma&cal Func&on Assignment within the word:  MoroSUBJpronominalparadigm:OBJpronominalparadigm:

Pa\ern1(SM{1ST|2ND}‐)CMSUBJ3RD‐V‐ASP‐OMPa\ern2(SM{1ST|2ND}‐)CMSUBJ3RD‐OM‐V‐ASP

DedicatedSUBJandOBJposiDonsoccupiedbydedicatedpronominalSUBJandOBJmarkers.

1stand2ndSUBJalwayspronominal,3rdeitherpronominalorgenderagreementmarkerwithlexicalNPs;allOMspronominalanddonotreflectnounclassofnominal,unlike(oqen)inBantu.

1SG e-g-

2SG a-g-

3SG CM-

1INCL DUAL alə-g-

1PL INCL alə-g...r

I PL EXCL ɲa-g-

2PL ɲa-g-

3PL CM-

Pattern 1 Pattern 2PREFIX SUFFIX PREFIX SUFFIX

1SG -ɲe -ɲə-2SG -aŋa -ŋa-

3SG HUM -ŋo -ŋʷə-3SG NONHUM - -I INCL DUAL -nḍa -nḍə-1PL INCL -nḍ(-)r -nḍə- -r

1PL EXCL -alanḍa -ɲə- -alanḍa

2PL -nḍa -nḍə-

3PL -lo -lo

23

Recombinant parts and paSerns: Gramma&cal func&on marking within the verb (Ackerman & Rose in prepara&on)

Person Hierarchies recur in the morphologies of many languages:(Silverstein1976;A.EKibrik2003;Haspelmath2004;Siewierska2004;Corbe\2006,Cysouw2008,amongothers,but Bickel & Witzlack-Makarevich 2008 for critical evaluation.)

SeveralKordofanian(NubamountainsinSudan)languagesexhibitvariantsofacommonlya\estedPersonHierarchy:{1>2}>3

deployedtoindicateGrammaDcalfuncDonreversal.

Gramma7cal Func7on ReversalconcernstheassociaDonofCMandPNMslotswitheitherSUBJorOBJfuncDons,dependingonco‐occurringfeaturesetsasmediatedbytonewithintheword.

Cwaya, Otoro, Tira, Koalib:morphologicalmarkersareassociatedwithSUBJorOBJstatusdependingonthepersonnumbervaluesassociatedincorporatedpronounwhen1&2ndco‐occurwith3rdorwhentwo3rdpersonsco‐occur.

24

Gramma&cal Func&on Reversal  (Klaiman1992forgeneraoverview;Samvelian2007&BonamiandSamvelian2009onSoraniKurdish;Baerman2007onNeo‐AramaicandInverse Markinginseverallanguages)

InSoraniKurdish,reversalscorrelatewithspecifictensevalues,butapersonoranimacyhierarchyisnotreportedtoberelevant.

a.(min)boNamin=îda‐kir‐im(I)forNarmin3.sgam‐buy.pres‐1.sg`IambuyingitforNarmin’

b.(min)kiteb=imboNamînkir‐î(I)book1.sgforNarminbuy.past‐3sg`IboughtabookforNarmin’

c.badûrbi=mândit‐inwithbinoculars‐1.plsee.past‐3.pl`Wesawthemwithbionoculars’.

2ndposiDoncliDcisassociatedwithanOBJpronominalandtheverbalsuffixisassociatedwiththeSUBJin(a),butthesegrammaDcalrolesarereversedin(b)in(c).

25

Gramma&cal Func&on Reversal in Cwaya: Simplified  (Samvelian2007&BonamiandSamvelian2009onSoraniKurdish;Baerman2007onNeo‐AramaicandInverse Markinginseverallanguages)

TransiDveverbsinCwaya:(basedonGuest1998;Stevenson/SchadebergonTiraandOtoro2009;SchadebergandKossmann2010;Quint2006ontheroleoftonalmelodiesforreversalinKoalib.)

Partial Pattern 1 CMSUBJ/OBJ-VTRANS-{PNMSUBJ/OBJ-3PLSUBJ} Partial Pattern 2 CMSUBJ/OBJ-TNS-{PNMSUBJ/OBJ-3PLSUBJ}-VTRANS

CM=3rdperson,PNM =pronounsfor3persons/2numbers,{ } =variableposiDonsubtemplatewith2slots.

BothCMandPNMcanbeSUBJorOBJwithmulDpleexponenceforpluralityofSUBJ.

ObservaDons:

CMisnotadedicatedSUBJmarkerandthevariableslotsintheverbareassociatedwithPNMsarenotdedicatedOBJs‐unlikeMoro.

PosiDonofPNMwithintheverbtemplate,likeMoro,is(partly)anexponentof(setsof)morphosyntacDcpropertysets.

26

When1stand2ndco‐occur,thenthereisonlyasingleincorporatedpronounanditistheOBJ:CM funcDonsasnumberagreementmarker.

        1.  nyixa‐má‐nga‐ridi1SG.SUBJSG.SUBJ‐PAST‐2SG.OBJ‐play`Iplayedwithyou’

2.ngaxa‐má‐nyí‐rídi2SG.SUBJSG.SUBJ‐PAST‐1SG.OBJ‐play`Youplayedwithme’

Independentpronounco‐occurswithCMdisplayingnumberagreementandwhenSUBJisplural,then‐llaappearsaqerOM:

3.ánángála‐m‐á‐lla‐rahm1pl.SUBJPL.SUBJ‐PAST‐2SG.OBJ‐PL.SUBJ‐bite`Webityou’

27

1st and 2nd person

When1or2co‐occurswith3rd,thenbothareincorporatedintoverb,butgrammaDcalfuncDonisdeterminedbytone,whilesegmentalmarkers(generally)remainthesameirrespecDveoffuncDon.

CM isa3rdSUBJexhibiDnggenderclassandPNMis1STOR2NDOBJ:HIGHTONE

Pasttensevowelishigh:nga‐ma‐ny‐apáCMSUBJ‐PNMOBJ‐V3SG‐PAST.1SG.carry`Itcarriedme‘

CMisa3OBJandPNMis1STOR2NDOBJ:LOWTONE

Pasttensevowelislow:nga‐ma‐ny‐apáCMOBJ‐PNMSUBJ‐V3SG.PAST.1SGJ.carry`Icarriedit‘

(ngahereisagenderclassmarkerthatissegmentallysyncreDcwiththe2.sg

incorporatedpronoun.)

28

{1,2} person and 3rd

Summary

GrammaDcalfuncDonstatusofCMsandPNMsreflectsa1&2versus3rdopposiDon;

Manyofthesamepiecesarere‐usedfordifferentpurposescorrespondingtowordswithdifferentfuncDonassignments,i.e.,theformngacanbeasingular3rdagreementmarker,a3rdsingularincorporatedSUBJ orOBJ.  

Cross‐linguisDcallyperson‐hierarchiesplayaroleinvariousmorphologicalphenomena,languageshaveSUBJandOBJpronounincorporaDon,languageshavevariableaffixorders,languageshavegrammaDcaltone,person‐numbercanbeexpresseddisconDnuously,therecanbemulDpleexponence,so,inprinciplethereproperDescancombineinnovelways.

29

Summary: Moro and Cwaya

Moro: Pa\ern1(SM{1ST|2ND}‐)CMSUBJ3RD‐V‐ASP‐OMPa\ern2(SM{1ST|2ND}‐)CMSUBJ3RD‐OM‐V‐ASP

Cwaya: Pa\ern1CMSUBJ/OBJ‐VTRANS‐{PNMSUBJ/OBJ‐3PLSUBJ}Pa\ern2CMSUBJ/OBJ‐TNS‐{PNMSUBJ/OBJ‐3PLSUBJ}‐VTRANS

Dothesespecificpa\ernsfollowfromprinciplesofgrammararchitectureorratherfromthedynamicsofhistoricalconDngencieswithin(related)grammarsystemsasconstrainedbyprinciplesofhumancogniDon,andparadigmorganizaDon?

Wedon’thaveenoughcomparaDvedataonKordofaniantoidenDfythespecificsystemicexplanaDon(s)forthesimilariDesanddifferencesatplayhere.

But,wedohaveenoughcomparaDvedatatomoDvate/explainadifferent“anomalous”phenomenon,namelyPossessive Rela7ve Construc7ons.

30

3. Systemic explana&on of the possible: Poten&a&ng the possible (Ackerman & Nikolaeva Descriptive Typology and Grammatical Theory To appear)

Empiricalproblem:ManygeneDcallyrelatedandunrelatedlanguagesinEurasiacontainanunusualrelaDveclauseconstrucDon.

ItisinstrucDveaboutthenatureofgrammarorganizaDon,andhence,thenatureofgrammaDcalarchitecturemorebroadlyconstrued.

ThemostfelicitousdescripDonaswellasthemostinsigh�ulexplanaDon(moDvaDon)forthisphenomenonsuggestsapa\ern‐theoreDcapproachtogrammaranalysis.

SimilarinspirittoHarris’(2007)argumentthatsystemicproperDesofgrammarslicense‘odd’construcDonssuchasUdiendoclisisandGeorgiancasemarking.

Ratherthanlookingatasinglephenomenoninalonelanguage,weanalyzean‘odd’construcDontypethatappearsinnumerousrelatedandunrelatedlanguages.

31

Prenominal rela&ve clauses

ManylanguageshaveexternallyheadedprenominalNON‐SUBJECTrelaDves:

[[øGAP…VMC]LOCALDOMAINHN]EXTERNALDOMAIN

NON‐SUBJNON‐SUBJ

builthouse

`thebuilthouse’

Diagram1

1.TherelaDvefuncDonsasthemodifieroftherelaDvizedheadnominal(HN)2.Thelocaldomainheadedbytheverbalmixedcategory(MC)isafullclause3.TherelaDvizednominalbearsaNON‐SUBJECT,(OBJ,ADJUNCT...)relaDontothegap4.GapsimplyaconvenDonforindicaDngthatsomethingismissinginthelocaldomainthatbearsasyntacDc&semanDcrelaDontotheVMC.

Q:HowisapronominalSUBJexpressed?

32

PaSern 1

M(ixed)C(ategory)‐inflectedrelaDve:person‐numbermarker(PNM)expressesSUBJpronominalontheVMIXEDCATEGORY

[[øGAP...VMC‐PNMSUBJ]LOCALDOMAINHN]EXTERNALDOMAIN

Diagram2

EasternOstyak(Uralic):

[[øGAP…wer‐t‐äm]LOCALDOMAINkiriw]EXTERNALDOMAIN

MAKEMC‐1SG.SUBJBOAT`theboatIwillmake’

Diagram3

ThePNMislocal tothedomaindefinedbytheverb.

33

PaSern 2

Possessive rela7ve‐person‐numbermarking(PNM)expressesSUBJpronominalontheHN:[[øGAP...VMC]LOCALDOMAINHN‐PNMSUBJ]EXTERNALDOMAINN

TundraNenets(Uralic):

[[øGAP…ta‐wi°]LOCALDOMAINte‐da]EXTERNALDOMAINgiveMCreindeer‐3SG‘thereindeerhe/shegave’

Observa&on 1:ThePNMseemstobeinthewrongplace,i.e.,itbearsaSUBJrelaDontotheVMCheadingthemodifyingclause.(runsafouloflocality)

Nominal Possessive Construc7ons:Head‐marking strategy   

serakote‐dawhitereindeer‐3sg`his/herreindeer’

Q:IstheresemblancebetweentheseindependentconstrucDonsfortuitous?

34

Possessive rela&ve clauses

Ques&on 1:DothesedistribuDonsfollowfromanytheory,i.e.,aretheypredicted?

Observa?on 2:  Every theory can deploy its tools to redescribe these distribu7ons: this is a minimal condi7on of adequacy for analysis. 

Basic Challenge:Is there a way to mo7vate/explain why the Possessive Rela7ve looks the way it does and is reliably iden7cal to nominal possessives wherever it occurs? 

35

PaSern‐theore&c gambit:  Dynamic systems explana&on

Guiding intui&on:AlllanguageswithPossessiveRelaDves(PRC)containthesamefourindependentlicensingconstrucDons

PossessedNounMorphologicalCX

Modifier‐HeadSyntacDcCX

InflectableNon‐finiteVMorphologicalCX

PossessiveRelaDveConstrucDon

36

Non‐finiteVSyntacDcCX

PaSern‐theore&c gambit:  Dynamic systems explana&on

Recombina7on of various elements found in independent morphological and syntac7c construc7ons cooperate to probabilis7cally yield a systemic redeployment of a new configura7on, the Possessive Rela7ve.

Pr(c1,c2,c3,c4|PRC)≈1IfalanguagehasPRC,itispotenDatedbyc1‐c4.

Mongolic:Kalmyk,Dagur,KhalkaMongolian,Buriat;Turkic:Altai,Uzbek,Turkmen,Tuva,Shor;Tungus:Evenki;Uralic:Nganasan,Enets,Vogul,Mari;IE:Western Armenian;Isolate:Yukaghir

Pr(PRC|c1,c2,c3,c4)=?Ifalanguagehasc1‐c4,can’tpredictpresenceofPRC.

Whatwewantaredisconfirmingdata.

37

Summary

HypothesisthatgrammarisacomplexsysteminwhichinteracDonsbetweenitsmanydimensionsandtheiringredientsproduceacanalizingordirecDnginfluenceconcerningwhatsortsofgrammarproperDesandconstrucDonsmayariseoverDme.

PotenDaDnginfluenceratherthanadeterminisDconesinceinmanyinstancesstructurespermi\edbyparDcularinteracDonssimplydonotoccur,thoughtheycouldhave,givendifferentconDngentcondiDons.

Certainstructurespossessanexceedinglylowprobabilityofarising,sincesystemicinteracDonsareunlikelytoproducethem.(cf.Harris2007)

Hypothesis:AseriesofconDngent,systemicpathways,someDmesguidedbyanalogy,togetherwithhumancogniDvecapaciDesprobabilisDcallydetermineobservedoutcomes.

Q:WhatprinciplesmayguidetheorganizaDonofsyntacDcandmorphologicalsystemsandwhatisthenatureoftheresulDngorganizaDon?

Don’tknowfortheorganizaDonalsystemforsyntax,butwehavesomeideasaboutmorphology.

38

4. Measuring the organiza&on of paradigm systems (Cherry, Halle, and Jakobson 1953; Moscoso del Prado Marmn et al. 2004; Milin et al. 2009a, 2009b, D. Brown 2010; Surrey Morphological Complexity Project)

We’veseenhowcomplexmorphologicalsystems(Cwaya)canberecombinantcombinaDonsoffamiliarandhowtheirinteracDonwithsyntax(TundraNenets)canproduceunusualsynchronicpa\ernsthatyieldtosystemicexplanaDon.

Thisseemstosuggestalotoflearningonthepartofspeakers,buthowcouldlearningofsuchcomplexsystemsoccur?

Howcouldcomplexmorphologicalsystemsbeeasytolearn?(Ramscar&Dye2010,Chater&ChrisDansen2010;Ramscar2011;

39

The Paradigm Cell Filling Problem (Paul 1890/1970; Paunonen 1974; Anqla 

1977; Bybee 1985; Thyme 1989;  Itkonen 2005, among others) 

SpeakersoflanguageswithcomplexmorphologyandmulDpleinflecDonclassesmustgeneralizebeyonddirectexperience,sinceit’simplausibletoimaginetheywillhaveencounteredeachformofeveryword

Paradigm Cell Filling Problem:Givenexposuretoaninflectedwordformofanovellexeme,whatlicensesreliableinferencesabouttheotherwordformsinitsinflecDonalfamily?(Ackerman,Blevins,&Malouf2009;Bonamiet.al.2010)

Morphologicalsystemsarenotunstructuredinventoriesbutexhibitpa\ernsofinterdependencebetweenforms.

Thesepa\ernsaretradiDonallyexpressedintermsofimplicaDonalorpredicDverelaDonsbetweenelements.

Apa\erninvolvingelementsAandBcanbedescribedintermsoftheimplicaDonalrelaDonsbetweenAandB.

The basic background

InflecDonalmorphologycanexhibitspectacularcomplexityin:

i. syntagmaDc,morphophonemic,suprasegmentalstructureofindividualwords;

ii. thesizeofinventoriesformorphosyntacDcdisDncDonsformallexpressedbywords;

iii. paradigmaDcpa\ernsthat(classesof)wordsparDcipatein.

ThisistheExternalComplexityorE‐complexityofamorphologicalsystem

Sixcasesandstemallomorphywithweakandstrongconsonantgrades

Our guiding intui&on

MorphologicalsystemsmustbesimpleinwaysthatallowthemtobelearnedandusedbynaDvespeakers,irrespecDveofhowcomplexwordsandparadigmsmayappearaccordingtoexternalmeasures.

Morphologicalsystemsmustpermitspeakerstomakeaccurateguessesaboutunknownformsoflexemesbasedonknownforms.

WordsparDcipateinimplicaDonalrelaDonswithrelatedwordsandthisnarrowsthesetofcandidatepossibleformsassociatedwiththeknownformsofalexeme.

ThisistheInternal SimplicityorI‐simplicityofasystem

Informa&on‐theore&c measures: the intui&on

ImplicaDonscanbemodelledbyuncertaintyreducDon:

1)AnelementAimpliesB(orBisdeduciblefromA)totheextentthatknowledgeofAreducesuncertaintyaboutB.

2)TheentropyofB,H(B),representstheuncertaintyassociatedwithB.

3)ThecondiDonalentropyofBgivenA,H(B|A)representstheremaininguncertaintygivenknowledgeofA.

CondiDonalentropyprovidesameasureofhowdifficultitistoreliablyguesse.g.,oneformofawordgivenanother:

thelowertheentropy,themoreconfidentonecanbeabouttheformofanunknownword,with0condiDonalentropyrepresenDngcompleteuncertaintyreducDon.

Our hypothesis: I‐simplicity

I‐simplicityismeasurableandquanDfiable

Principle of Low Paradigm Entropy:ParadigmstendtohavelowexpectedcondiDonalentropy,whereParadigmentropyistheaverageofcondiDonalentropiesamongallpairsofwords.

GradaDoninfirstdeclensionnounsinSaami(Bartens1989:511)

GivenaNOMINALSINGULARformhoweasyisittoguesstheLOCATIVEPLURALform?ThelowerthecondiDonalentropy,themorecertainonecanbeaboutatargetform.

Maximally Transparent Paradigms (Stump & Finkal 2007, To 

Appear)

Set‐theoreDcally:LexemeLismaximally transparentinparadigmP,ifanycellinPcanserveasL’ssufficientformtopredicttherestofitsinflectedforms.(AdaptedfromStumpandFinkal2007)

InformaDon‐theoreDcally:ifyouknowanysingleword,thereisnosurpriseassociatedwithidenDfyingthepreviouslyunencounteredformsofwordsorproducingatargetform.

45

Results based on equiprobable type frequencies (Malouf and 

Ackerman 2010)

Language Declensions Cells Realiza&ons Paradigm entropy

Bootstap Avg

Bootstrap p

Arapesh 26 2 41 0.630 0.630 1.000Burmeso 2 12 24 0.000 0.000 1.000Fur 19 12 80 0.517 1.316 0.001Kwerba 4 12 26 0.428 0.523 0.001NgiD 10 16 68 0.484 1.019 0.001Nuer 16 6 12 0.793 0.811 0.160Russian 4 12 26 0.538 0.541 0.383

Fur

Summary

WhiledetaileddescripDveaccountsofmorphologicalsystemsfoundingrammarscanprovideausefulentry pointforanalysis,providinganupperboundonentropycalculaDons,theycanalsoproducemisleadingconclusionsaboutParadigmEntropy(cf.Bonamiet.al.2011andSims2011).

ExploringthevalidityoftheLowEntropyConjecturerequires:

1)veridicalrepresenta&onsforwordformsinordertoreflecttheactualsDmuliencounteredbyspeakers(wefailedinourefforttomodelTundraNenetsinthisrespect),aswellas,

2)accuratetypeandtokenfrequencies, sinceequiprobabilityassumpDons,i.e.,thatthelikelihoodofencounteringallpa\ernsisthesame,areunrealisDc.

ParaphrasingBonamiet.al.(2011):thisis“tedious work”,butit’sbothdoableandnecessaryifwereallywanttounderstandmorphologicalsystems.

FollowingstandardinvesDgaDveprocedures,wehavetolookwherewe’reliabletobewrong,ratherthanwherewesuspectthatwe’llberight.

Tedious work: Modern Irish paradigm entropy (Malouf and 

Ackerman 2011)

Refiningthehypothesis:

1,200fullydeclinedIrishnounsbasedonCarnie2008andreflecDngtranscripDonoforthographicconvenDonstomoreveridicalphoneDcrepresentaDon:

Expandingthedatabaseto10,000fullydeclinednounswithtypeandtokenfrequencyinformaDon.

H(COL|ROW)COM.SG GEN.SG PREP.SG COM.PL GEN.PL PREP.PL E[ROW]COM.SG 1.003 0.808 0.976 0.104 1.011 0.780GEN.SG 0.723 0.840 0.039 0.602 0.010 0.443PREP.SG 0.304 0.617 0.594 0.110 0.622 0.449COM.PL 0.770 0.113 0.892 0.603 0.123 0.500GEN.PL 0.467 1.245 0.976 1.172 1.250 1.022PREP.PL 0.724 0.003 0.838 0.041 0.600 0.441E[COL] 0.598 0.596 0.871 0.565 0.404 0.603 0.606

Ques&ons being researched by this approach

1. Howarewordsorganizedintopa\ernswithinamorphologicalsystem?

2. HowcanoneidenDfyimplicaDverelaDonsbetweenwords?

3. HowmighttheimplicaDveorganizaDonofasystemcontributetolicensinginferencesthatsolvetheparadigmcellfillingproblem?

4. HowdoesthisorganizaDon,andthesurfaceinferencesitlicenses,contributetotherobustnessandlearnabilityofcomplexmorphologicalsystems?

5.  Concluding observa&ons: Morphology as a complex adap&ve system

“Inplaceofexplicitlycodingforapa\ernbymeansofablueprintorrecipe,self‐organizedpa\ernformaDonreliesonposiDvefeedback,negaDvefeedback,andadynamicsysteminvolvinglargenumbersofacDonsandinteracDons...environmentalrandomnesscanactasthe`imagina&on of the system’,therawmaterialfromwhichstructuresarise.FluctuaDonscanactasseedsfromwhichpa\ernsandstructuresarenucleatedandgrow.Theprecisepa\ernsthatemergeareoqentheresultofnegaDvefeedbackprovidedbytheserandomfeaturesofenvironmentandthephysicalconstraintstheyimpose,notbybehaviorsexplicitlycodedwithintheindividual’sgenome.”Camazine2001:26

Moro morphotac7cs, Cwaya gramma7cal func7on reversal, Tundra Nenets rela7ve clauses, and Estonian nominal declension paradigms arise in the “imagina7on” of their respec7ve grammar systems which recombine cross‐linguis7cally familiar ingredients into some7mes familiar and some7mes novel pa_erns in ways that make them learnable.

50

Alignments with big issues (adap&ng S&les 2009 for developmental psychology)

EvoluDonaryPsychology/(Neuro‐)construcDvist/DevelopmentalBiology/DevelopmentalPsychobiology/DevelopmentPsychology/(Ecological)EvoluDonaryMainstreamGeneraDveDevelopmentalBiology/GrammarPa\erntheoreDcgrammars/InformaDon‐theoreDcgrammars

Issue Nativist Developmental Science

Roleofexperience:coreconceptsversusacquiredconcepts

Core concepts consDtute a small butessenDal subset of constructs. Coreconcepts are present in the absence ofdirectexperience.

Conceptsdevelop from the interacDon ofbasic sensory and motor abiliDes andexperience with the world. There are nocoreconcepts.

Domainspecificversusdomaingeneral(innateversusacquiredmodularity)

Core concepts are domain‐specific andencapsulated from other imformaDonsources(i.e., theyaremodular fromthestart).

Domain general learning mechanismsunderl ie conceptual development.Modularity of systems is the normalproductofdevelopment,notitsstartstate.

InvarianceOnemarkofacoreconceptisthat itisconstantoveraspanofdevelopment.

Because knowledge is emergent andconstructed by thechild, change isacrossdevelopment.

UniversalityCore concepts are constant acrosscultures.

Ub iqu i ty does not enta i l i nnateoriginaDon. AdaptaDon to universalcond iDons can produce commonconstructs.

TriggeringversusinducDon

Environmental inputs serve to “trigger”theavailabilityofcoreconcepts.

Concepts are acquired and refinedthrough inducDon, i.e., hypothesisformaDonandtesDng

51