Realizing morphemes in the Icelandic noun phrase Anton Karl Ingason University of Pennsylvania. Contents

Transcription

1 Realizig morphemes i the Iceladic ou phrase Ato Karl Igaso Uiversity of Pesylvaia Dissertatio Proposal 9 October, 2014 Dissertatio supervisor: David Embick Proposal committee: Athoy Kroch, Julie Ae Legate, Floria Schwarz Cotets 1 Itroductio 2 2 Locality i Distributed Morphology Liear adjacecy ad cotextual allomorphy Phase locality Iterveers for affixatio Sectio summary Part I: Applicatives i the ou phrase The Iceladic Caused Experiece costructio The category head closest to the root Applicatives are idepedet of verbal morphosytax A causative omializer Sectio summary Part II: Two defiite articles Weak ad strog defiite articles Iceladic the-support Idepedet diagostics Sectio summary Part III: Further case studies The lack of idiosycratic declesio i Iceladic ous Phase locality ad two types of aget omials Iceladic umlaut ad strig phoology Sectio summary Theoretical ad empirical impact 36 7 Timelie 37 1

2 1 Itroductio Liguistics has subfields like sytax, phoology ad sematics ad puzzles of liguistic aalysis are commoly framed withi such fields. 1 It is well motivated that these subfields reflect somethig real about the orgaizatio of laguage i the brai. For example, it is reasoable to thik that the mechaisms that assemble hierarchical structure i sytax are qualitatively differet from the mechaisms that arbitrate segmetal phoology. However, it is also clear that these compoets are itercoected ad the coectios have cosequeces. This proposed dissertatio studies pheomea where the aalysis of liguistic soud objects is oly well iformed if it cosiders evidece of sytactic structure ad its iterpretatio i cotext. I such cases, the sytax makes predictios about the ways i which morphemes ca ad caot be realized. Importatly, if we uderstad how surface patters i morphology ad meaig are related to the sytax, we are i a positio where surface observatios about morphemes iform ad costrai sytactic aalysis. Theoretically, I aim to defed a strog versio of the view that surface complexity i atural laguage is the result of a iteractio betwee differet liguistic systems: The theory of realizatio of morphemes imposes boudaries o sytactic aalysis. I order to idetify ad aalyze morphemes, we eed to examie their realizatio i both prouciatio ad iterpretatio. These broad goals ca be addressed rigorously i a settig where a precise theoretical framework meets a well defied empirical case i which may differet liguistic costraits iteract with each other. The theoretical foudatio of the dissertatio will be the framework of Distributed Morphology (DM, Halle ad Maratz 1993 ad subsequet work) ad I will ivestigate the realizatio of Iceladic ous, icludig their roots, omializers, iflectio morphemes ad markers of defiiteess, all of which may appear i the same ou, as show i (1). (1) leik-ed-ur-ir play-omlz-masc.om.pl-def the actors Each morpheme raises questios about realizatio i cotext: To what extet ca the idetity of the root ifluece the prouciatio of the other morphemes ad vice versa? How does derivatioal morphology restrict the iterpretatio of the root? Is there ay uproouced structure i the word ad how does it affect the surface form? Are there limits to the variatio associated with the so-called declesio classes which are reflected i the expressio of geder, case ad umber? Due to which kid of a mechaism is the defiite article i Iceladic usually a suffix, but ot always? Research i DM has provided evidece for locality priciples which costrai the realizatio of morphemes at the iterfaces of sytax (Embick 2010, Maratz 2013). O the oe had, the prouciatio ad iterpretatio of a morpheme is sesitive to deep sytactic mechaisms like phase locality, which is a cosequece of cyclic Spell-Out i sytax (Chomsky 2001). Thus, the soud form of a morpheme ca oly be determied by material which is phase-theoretically active at the poit i the derivatio at which the allomorph selectio takes place. Furthermore, phase heads are the locus of polysemy resolutio i the iterpretatio compoet ad therefore phase structure limits the rage of possible iterpretatios. O the other had, the realizatio of the same morphemes is also costraied by more shallow iterface-specific mechaisms which are sesitive to otios like liear adjacecy. Therefore, coditios o the form of a morpheme require the coditioig eviromet to be ot oly phase-active but also liearly adjacet to the positio which is beig realized. This liear effect is shallow because there is o pricipled reaso for sytax to care about liear order. However, the liear ature of time meas that we proouce morphemes oe after the other, ad it appears to be a fact that speakers ca select the form of a morpheme based o properties of the precedig ad the followig morphemes. We refer to such selectio as cotextual allomorphy. 1 I use the followig abbriviatios: acc(usative), dat(ative), def(iite), fem(imie), ge(itive), lm (likig morpheme), masc(ulie), eut(er), om(iative), omlz (omializer), pl(ural), refl(exive), sg (sigular). 2

3 The Iceladic ous i (2) are represetative of the relevat types of empirical distictios. They are all built o the root leik play but they arguably differ with respect to the structure of their category heads. Category heads are assumed to be phase heads ad they iclude (omializers) ad v(erbalizers) (Maratz 2001, 2007). The label Ifl stads for the omial iflectio morpheme which expresses geder, case ad umber. (2) a. leikur game b. leikari actor c. leikadi actor leik leik Ifl -ur leik leik -ar Ifl -i leik leik v v -ad Ifl -i The differece betwee (2a) ad (2b) is that the latter cotais overt omializatio morphology but the former does ot. This meas that Ifl is liearly adjacet to the root i leik-ur game but ot i leik-ar-i actor. The differece betwee (2b) ad (2c) is that the latter cotais a verbal layer which is ot preset i the former (Igaso ad Sigurðsso 2014 [forthcomig]). This meas that despite superficial similarities, the omializer is phase-local to the root i leik-ar-i actor but ot i leik-ad-i actor. These differeces have cosequeces that will be elaborated i the followig sectio ad throughout this dissertatio proposal. The poit to be made right away is that a well iformed aalysis of the soud form has to ask questios like the followig, assumig that a morpheme geerally represets a piece of meaig. Which pieces of meaig are there i the word? Which of these pieces have a overt phoological expoet? How may category heads are there i the word, whether overt or covert? Empirically, I focus o the Iceladic ou phrase as well as comparative evidece from Germaic. A wide rage of locality effects is attested i Iceladic ous. This is pleasig because it shows that the locality priciples are ot artifacts of coveiet but cross-liguistically isolated data sets. Several locality effects ca be maifested i oe Iceladic ou. The expositio is orgaized aroud two large ad a few smaller case studies. Part I: Applicatives i the ou phrase: Cotrary to the predictios of the stadard theory of applicatives (Pylkkäae 2008), Iceladic ous are show to itroduce dative experiecers without havig a verbal substructure. The study compares Eglish geruds ad a Iceladic Caused Experiece costructio. The crucial evidece ivolves costraits o cotextual allomorphy ad polysemy resolutio. Part II: Two defiite articles: This comparative study of Germa ad Iceladic supports the proposal by Schwarz (2009) that there exist two types of defiite articles, a weak uiqueess article ad a strog aaphoric article. I Iceladic, the uiqueess article participates i support morphology but the aaphoric article does ot. The support pheomeo, dubbed the-support, is structurally similar to Eglish do-support ad has implicatios for the role of liear adjacecy ad head-to-head relatios i affixatio (cf. Embick ad Noyer 2001). Part III: Further case studies: I preset (i) a large scale empirical study o root adjacecy ad Ifl i (2a) vs. (2b), (ii) a study o phase locality i aget omials of types (2b) vs. (2c), which argues that oe eeds to uderstad sytax ad sematics i order to aalyze Iceladic palatalizatio, ad (iii) a study o the status of adjacecy i the applicatio of the Iceladic umlaut. I compare ad cotrast the pheomea i these smaller studies to illustrate differet ways i which phoology iteracts with morphology. 3

4 This proposal is orgaized as follows. A brief theoretical backgroud is give i 2. The case study o applicatives is preseted i 3, the two defiite articles i 4 ad the smaller studies i 5. I summarize the theoretical ad empirical implicatios of the proposed dissertatio i 6. 2 Locality i Distributed Morphology The cetral theoretical basis for the dissertatio is the part of Distributed Morphology which imposes limits o how morphemes ca be realized. This sectio gives a brief theoretical backgroud. I first review the role of liear adjacecy ad phase locality i cotextual allomorphy ad meaig ad the I discuss the role of iterveers for affixatio i pheomea like Eglish do-support. 2.1 Liear adjacecy ad cotextual allomorphy I follow Embick (2010) i assumig that coditios o cotextual allomorphy ca oly refer to liearly adjacet morphemes. This adjacecy costrait is ot a primitive otio i the framework but rather a effect which is derived from the fact that the mechaism which arbitrates cotextual allomorphy is stated i terms of cocateatio. Cosider the two Iceladic ous i (3). They are both built o the root leik play ad they crucially differ with respect to whether they cotai overt omializatio morphology or ot. (3) a. leikur game b. leikari actor leik leik Ifl -ur leik leik -ar Ifl -i I DM, the phoological expoet of a morpheme is determied by a set of ordered Vocabulary Items. A toy aalysis for the omiative sigular of the above words is give below. The Vocabulary Items state that Ifl is proouced -ur if it is cocateated with oe of the roots o the list but otherwise -i is iserted. 2 (4) a. Ifl[masc, om, sg] -ur / { leik, bekk,...} _ b. Ifl[masc, om, sg] -i Cosider the derivatio of the root-adjacet Ifl -ur i leik-ur actor i (6). The sytactic structure (5a) is coverted to a set of cocateatio statemets (5b). Vocabulary Isertio takes place from the iside out (see Bobaljik 2000), first by makig - the phoological expoet of (5c). Null odes are removed by a operatio of Pruig which results i Ifl beig cocateated with the root (5d). The Vocabulary Items i (4) state that Ifl is realized as -ur whe cocateated with the root leik as is show (5e). Fially, everythig is chaied together (5f). (5) a. leik Ifl [masc,om,sg] b. c. leik, Ifl leik [,- ], [,- ] Ifl (Cocateatio) (Vocabulary Isertio) 2 The example is costructed to demostrate the adjacecy costrait ad it igores various aspects of these morphemes, icludig potetial for uderspecified Vocabulary Items. 4

5 d. e. f. leik Ifl leik [Ifl,-ur] leik-ur (Pruig) (Vocabulary Isertio) (Chaiig) The derivatio of leik-ur game shows how the Pruig of ull puts the root withi reach of the mechaism that determies the form of Ifl. We ca ow cotrast the above with leik-ar-i actor i which the morphology is overt. The sytax is the same (6a) ad after cocateatio (6b), -ar is iserted as the form of agetive (6c). There is o ull morpheme here, so o Pruig takes place, ad accordig to the rules i (4), Ifl is realized as -i (6d). (6) a. leik agt Ifl [masc,om,sg] b. c. d. e. leik agt, agt Ifl leik [ agt,-ar], [ agt,-ar] Ifl leik [ agt,-ar], [ agt,-ar] [Ifl,-i] leik-ar-i (Cocateatio) (Vocabulary Isertio) (Vocabulary Isertio) (Chaiig) Coditios o cotextual allomorphy are expressed i terms of the cocateatio operator, as i (4a), ad therefore it caot see past overt iterveig material. I additio to the adjacecy costrait, the preset system assumes that phase-theoretic locality costrais cotextual allomorphy. 2.2 Phase locality I assume that sytactic structures are set cyclically to the iterfaces, by phase (Chomsky 2000, 2001), for phoology (PF) ad iterpretatio (LF), ad that the cyclic ature of the derivatio imposes locality costraits. Although cosiderable work has bee doe i recastig earlier work i sytax i terms of phases, particularly with respect to costraits o movemet, there is also evidece that phases play a role i surface pheomea at the iterfaces. My formulatio of such locality effects mostly follows the implemetatio i Embick (2010, 2013a), which assumes the secod versio of Chomsky s Phase Impeetrability Coditio (PIC2). (7) Phases Category-defiig heads, at least v,, a (Maratz 2001, 2007), ad C, ad possibly others, trigger Spell-Out of their complemets; they are the phase heads. 3 These phase heads are ofte realized as derivatioal morphemes. The cyclic Spell-Out meas that oly a sub-part of the structure is visible (or active) at ay give poit. I this type of a theory, every sytactic termial ode is a morpheme, ad cyclicity costrais iteractio betwee morphemes. (8) Phase locality A phase head caot see morphemes across the ext phase head. A o-phase-head ca see morphemes across (just) the ext phase head. 3 For the purpose of sytactic displacemet, C is the most caoical phase head. Depedig o other assumptios, the category v correspods for the most part to v i a o-dm miimalist framework, although a sythesis of the differet implemetatios of phase theory is a o-trivial theoretical matter. It is less obvious i the sytax that all istaces of v are phases, but I will evertheless assume so. O every v beig a phase, see Legate (2003); Igaso ad Wood (2013, 2014 [Submitted.]). 5

6 (9) P XP X vp v Y I (9), ad v are cyclic (phase) heads, whereas X ad Y are ot. Because of (8), sees X ad v but it does ot see Y. I cotrast, X sees v ad Y. Phases have cosequeces for realizatio of morphemes with respect to both morphology ad meaig. First, cosider morphology (Embick 2010, 2013a). (10) Phase locality ad morphology Coditios o allomorphy caot be stated i terms of ivisible (iactive) material. The differece betwee derived omials ad geruds i Eglish is a clear example of (10). I additio to a sharp split betwee the sytactic properties of the two types of ous, diverse allomorphy is attested i the case of derived omials whereas geruds always surface with the -ig suffix. (11) (derivied) omial gerud marri-age marry-ig destruct-io destroy-ig refus-al refus-ig cofus-io cofus-ig (12) root marri -age v root marry v -ig It is a well motivated sytactic aalysis to say that geruds, ulike derived omials, are verbs iterally, as reflected by the structural cotrast i (12) (cf., Chomsky 1970). This rederig of the structures is modeled after Embick (2010) (see also Maratz (1997)) where the of a derived omial combies directly with a root but i a gerud combies with a v which has already combied with a root. I the DM framework, the allomorphy cotrast is predicted because (10) meas that ca see a root that it combies with directly but it caot see a root across v. Root-coditioed allomorphy o i geruds is impossible. Moreover, the phase is also a locality domai for coditios o meaig. Whe arrowig dow the set of possible related meaigs for a root, a meaig that has bee excluded at a ier phase head is uavailable at a outer phase head. The phase heads ca therefore be viewed as the locus of subset operatios over possible meaigs of roots i polysemy resolutio (Maratz 2013, see also Arad 2003). (13) Phase locality ad meaig A meaig of a root that has bee excluded at a ier phase head is uavailable at a outer phase head. Cosider the polysemy of Eglish globe as a example (see Maratz 2013 for a more detailed discussio of the same example). (14) a. b. globe abstract sphere, somethig spherelike globe the world Oce the polysemy cloud of the root has bee restricted by a phase head (derivatioal morpheme), the excluded meaigs are uavailable to outer phases. At the root, the rage of possible meaigs icludes both of the above. Makig globe a ou allows for both meaigs, whereas the adjective global excludes the spherelike meaig. This cotrast is show i (15). 6

7 (15) Nou: spherelike world Adjective: * spherelike world a Adjective-derived: * spherelike world v globe - globe a -al a globe a -al v -ize As the rightmost example shows, (13) has the effect that the spherelike meaig that is excluded at the a phase head remais uavailable at the outer v phase head. Oce a possible meaig has bee excluded from the rage of optios, it caot be brought back. 2.3 Iterveers for affixatio Aother type of a locality effect i morphology is that some morphemes eed to attach to a immediately local morphological host. A well kow example of this type is do-support i Eglish. The T(ese) morpheme i Eglish eeds to be immediately local to v (the verb). I the absece of T/v iterveers, tese i Eglish is expressed as a suffix o the verb (16a). Negatio is a T/v iterveer ad therefore (16b) is ugrammatical; egatio triggers do-support (16c). However, vp-adjoied adverbs are ot T/v iterveers for the purpose of do-support (16d). The sytactic aalysis of (16c) is give i (17), based o Embick ad Noyer (2001:587), where do is a default v morpheme. (16) a. Joh agreed. b. * Joh ot agreed. c. Joh did ot agree. d. Joh completely agreed. (17) Sytax of Eglish do-support T T NegP v di T -d Neg ot vp agree While some otio of immediate locality is clearly eeded, it is a complicated matter to determie the most appropriate formulatio of this locality. For Eglish do-support, Embick ad Noyer propose a operatio of morphological lowerig i the phoological compoet that is costraied to a relatioship betwee a head ad the head of its complemet. The Eglish vp is ot the complemet of T whe egatio itervees ad therefore lowerig is blocked. Sice the relatioship is defied i terms of a head ad its complemet, a adverb that is adjoied to vp is correctly predicted ot to have a effect. A possible alterative is to say that the suffix T combies with its host, v, by so-called local dislocatio uder liear adjacecy (see Embick ad Noyer 2001). 4 I that case, a reaso has to be provided for why adjucts do ot cout as iterveers. There exist aalytical optios, e.g., some implemetatio of late adjuctio (Stepaov 2001). For the purpose of aalyzig Eglish do-support, it is ot obvious whether it is preferable to require the T morpheme to be liearly local to the verb or i a head-to-head selectioal relatioship with it. Either optio ca be reasoably etertaied, depedig o the theoretical prefereces of the aalyst. The dissertatio will iclude a case study o defiite articles i Iceladic (see 4) which has geeral cosequeces for the theory of iterveers for affixatio. 4 Local dislocatio ad lowerig build o earlier work o Morphological Merger (Maratz 1984, 1988). The relatioship betwee sytactic structures ad affixatio i Eglish was already a object of iquiry i Chomsky (1957). 7

8 2.4 Sectio summary This sectio provided a brief backgroud o the theory of locality i Distributed Morphology. This theory makes predictios about how deep sytactic mechaisms ad shallow iterface-specific mechaisms cospire to costrai liguistic surface forms. Three mai types of locality were idetified. Liear adjacecy: Morphemes ca coditio allomorphy o each other if they are adjacet to each other. Furthermore, it appears that affixatio of boud morphemes to a appropriate host does, i at least some cases, demad liear adjacecy betwee the affix ad the host. See 4 o adjacecy ad the affixatio of the Iceladic defiite article as well as related pheomea i Eglish do-support. See 5.1 o root-adjacecy i omial declesio. See 5.3 o the lack of adjacecy i the Iceladic umlaut. Phase locality: Sytactic structure is set cyclically to the iterfaces, by phase. Coditios o cotextual allomorphy require the coditioig eviromet to be phase-theoretically active. Phase heads are also the locus of subset operatios over sets of possible iterpretatios i polysemy resolutio. See 3 o phase locality i applicative structures. See 5.2 o phase locality i aget omials. Head-to-head locality: Head movemet i sytax is a displacemet relatioship betwee a head positio ad the head positio of its complemet. Affixatio may i some cases ivolve a morphological operatio of lowerig which is a mirror image of head movemet. If so, lowerig is structurally parallel to head movemet, but dowwards. See 4 o head-to-head locality i affixatio of the Iceladic defiite article ad Eglish do-support. All of these locality coditios play a importat role i the case studies which form the mai part of this dissertatio. The case studies are preseted i the followig sectios. 3 Part I: Applicatives i the ou phrase The first mai case study defeds the hypothesis that dative experiecers ad beefactives ca be itroduced iterally to omial structures that are ot verbs o the iside. 5 The discussio is framed with respect to the stadard theory of applicatives (Pylkkäe 2008) ad the fidigs are oteworthy because the theory eeds to be exteded i order to accommodate them. The crucial evidece ivolves phase locality i cotextual allomorphy ad polysemy resolutio ad therefore the case study shows how our uderstadig of surface patters i morphology ad sematics iform ad costrai sytactic aalysis. I first preset the core data ad the move o to a aalysis which will be further developed i the dissertatio. 3.1 The Iceladic Caused Experiece costructio The data i this case study come from the Iceladic Caused Experiece (CEx) costructio. Four varieties of the costructio are show (18 21). All the examples roughly mea that the dacig caused the girls to be etertaied. Note that the prepositio til for is part of (20 21) ad takes a 5 This case study is based o Igaso (2013a). Parts of this material were preseted at the 25th Scadiavia Coferece of Liguistics i Reykjavík, May 13, 2013, as well as at the Uiversity of Massachusetts, Amherst, April 26th, 2013, ad at the Uiversity of York, Oct 30,

9 geitive complemet as it ormally does i Iceladic. Here, the dative experiecer appears i Spec,P although it it ca also optioally follow the ou. 6 CEx, DP causig evet: (18) Dasi var stelpuum góð skemmtu. dace.the.om was girls.the.dat good etertaimet.om The dacig etertaied the girls well. CEx, PP causig evet: (19) Stelpuum var skemmtu *(af dasium). girls.the.dat was etertaimet.om *(by dace.the) The girls were well etertaied by the dacig. til-cex, Setece causig evet: (20) Þeir dösuðu [stelpuum til skemmtuar] they daced [girls.the.dat for etertaimet.ge] They daced for the girls etertaimet til-cex, DP causig evet: (21) Dasi var [stelpuum til skemmtuar]. dace.the.om was [girls.the.dat for etertaimet.ge] The dacig was for the girls etertaimet. The theory of applicatives (Pylkkäe 2008) is essetially a theory of how datives combie with verbs. It is impossible i that theory for a ou to take a dative experiecer argumet uless the ou is a verb o the iside like a Eglish gerud. I will argue that the dative experiecers i the CEx costructio are ot like geruds but they are evertheless itroduced withi the ou phrase. A few more represetative examples of CEx predicates are show i (22). (22) skemmtu etertaimet, léttir relief, hvatig ecouragemet, ydisauki pleasure, ejoymet, hressig refreshmet, dægradvöl recreatio, skaprau aoyace, botherig, vobrigði disappoitmet, iðurlægig humiliatio, álitshekkir reputatio damage Case sytax ad argumet structure i Iceladic clausal sytax have bee studied i great detail i a umber of works by several researchers throughout the last few decades (see Thráisso 1979; Zaee et al. 1985; Yip et al. 1987; Sigurðsso 1989; Jósso 1996; Thráisso 2007; Wood 2012 to ame a few). Iterestigly, much less attetio has bee paid to argumets i the Iceladic ou phrase 7 ad to my kowledge there is o previous systematic treatmet available of the dative experiecers i (18-21). I fact, previous work o Iceladic omializatios explicitly assumes that datives of this type caot be part of omializatios. I the approach take by Yip et al. (1987: ), the presumably lexical operatios of omializatios have, as oe compoet, the removal of lexical case, which rules out ou phrase iteral dative experieers i their system, ad Malig (2001: ) specifically rules out o-pp goal argumets i the Iceladic ou phrase. 8 The curret study shows that these geeralizatios are ot bore out by examiig facts that have escaped otice. The argumet is preseted from the iside out. I first show that the CEx predicate etertaimet is a ou iterally as well as exterally. I the show that the applied argumet the girls is itroduced withi the ou phrase, ad it is ot a part of the clause. A (base-geerated) aalysis is give i (23). 6 The obligatory by-phrase i (19) is retured to i 3.4 ad the costituecy idicated i (20 21) i There is a literature though that puts a primary focus o the orderig of elemets i the ou phrase, e.g. Magússo (1984); Sigurðsso (1993); Julie (2005). 8 The thematic role goal covers recipiets, experiecers ad beefactives i her termiology. 9

10 Its crucial property is that the closest category head to the root is. 9 This cause is realized with a a causative iterpretatio i the sematics. I retur to the causative aalysis i 3.4. I assume that the morphology of skemmtu etertaimet, o the right i (23), is derived by head movemet ad by attachig the omial iflectio Ifl. (23) PP P for cause -met P ApplP Appl root Appl cause Ifl DP the girls Appl exp etertai I fially commet o the distibutio of exteral eviromets i which CEx appears. If the applied dative is itroduced with oe of the CEx predicates, a causig evet has to be overtly expressed (as a setece, a DP subject, or a PP adjuct), ote the obligatory by-phrase i (19). A causative aalysis accouts for this distributio where cause deotes the causative sematics. 3.2 The category head closest to the root Sectio 2.2 reviewed costraits o cotextual allomorphy ad polysemy resolutio which are sesitive to phase structure. The costraits have the effect that ous which are verbs o the iside, like geruds, are more restricted tha other ous i terms of the realizatio of their morphemes because they ivolve a additioal Spell-Out cycle. The phase structure of derived omials ad geruds is give i (24) ad a few illustrative examples i (11), repeated as (25). (24) root v root v (25) (derivied) omial gerud marri-age marry-ig destruct-io destroy-ig refus-al refus-ig cofus-io cofus-ig There is ample evidece for the additioal verbal layer i geruds. These facts are well kow (e.g., Chomsky 1970) ad I will ot review them i detail here. They iclude compatibility with verbadjoied adverbs ad licesig of direct objects i the accusative case. For the preset purpose, the importat observatios are that (i) there always exists a verb which correspods to the gerud, (ii) the omializatio morphology is ivariably -ig i geruds whereas it is root-coditioed i derived omials, ad (iii) that the meaig of geruds correspods closely with the meaig of their iteral verb, icludig the aspect of meaig which determies thematic properties of the verb s argumets. Turig to Iceladic, we fid root-coditioed allomorphy i CEx ous ad this allomorphy is evidece that is the closest category head to the root. Represetative examples of CEx predicates 9 Uder a view i which a DM root phrase correspods to a big VP i o-dm Miimalism, the sytactic differece betwee this Appl aalysis ad the stadard theory is that Appl ca be selected by other category heads tha v, e.g., by. I am ot cocered with the implemetatio details of clausal applicatives here. The crucial facts here are that this dative is a applied argumet ad the closest category head to the root is omial. 10

11 are listed i (26) (om.sg. forms) ad they patter with derived omials rather tha geruds i their allomorphy. (26) Root-coditioed allomorphy i CEx: Root cause Ifl skemmt -u - etertaimet létt - -ir relief -ir - (alterative segmetatio of relief ) hvat -ig - ecouragemet ydisauk - -i pleasure hress -ig - refreshmet dægradvöl - - recreatio skaprau - - aoyace vobrigð -i - disappoitmet iðurlæg -ig - humiliatio álitshekk - -ir reputatio damage -ir - (alterative segmetatio of reputatio damage ) The cases of skemmt-u, hvat-ig, hress-ig, vobrigð-i ad ydisauk- -i show that the omializatio morphology ca be coditioed by the root i CEx ous. The same applies to létt-ir ad álitshekk-ir whe -ir is segmeted as a omializer by speakers, a o-stadard but widely attested patter. The cases which ivolve ull omializers furthermore show that the omial iflectio layer i CEx ca be coditioed by the root whe it is adjacet to the root. This is show by the differece betwee ydisauk-i ad the stadard segmetatio of létt-ir ad álitshekk-ir. While the weak declesio masculie -i for omiative sigular i ydisauk-i might be see as a default, -ir is associated with specific roots ad it is a rare patter. Accordig to the system i 2.2, the root-coditioed allomorphy is evidece agaist a verbal layer iside these ous. I will retur to the details of Ifl ad adjacecy with the root i 5.1. Note that Appl i the hypothesized structure i (27) is zero, so it will be Prued away. The root ad the omializer will therefore be cocateated with each other ad this allows them to coditio allomorphy o each other. Furthermore, because the theory assumes the secod versio of the Phase Impeetrability Coditio (PIC2), Ifl is realized while the root is still phase-theoretically active. Accordig to PIC2, Spell-Out does ot take place util the ext higher phase is merged. This meas that as log as is ot overt (for the purpose of liear adjacecy), Ifl ca be coditioed by the root i this structure. 10 (27) Ifl Appl root Appl cause We do furthermore ot fid regular verb/ou correspodeces of the Eglish gerud type. All of the followig types of mismatches are attested. For some CEx ous like ydisauki pleasure, lit. pleasure icrease, there is o correspodig verb made from the same matchig root material *ydisauka to pleasure. The most closely correspodig verb does ot always take a dative argumet, eve if the CEx ou does, as i the case of hvatig ecouragemet. The verb hvetja ecourage is a om-acc verb. 10 Although some Appl heads are aalyzed as phases by McGiis (2001), I do ot assume that Appl is a phase. 11

12 The most closely correspodig verb does ot always take a argumet which is iterpreted as a experiecer ad i some cases the meaig of the CEx costructio caot be paraphrased usig the verb. This lack of verb-ou correspodeces is summarized i the table below. The table shows that the mismatches are widespread ad do ot appear to be systematic. (28) CEx ous ad correspodig verbs. + idicates a ou/verb match. Paretheses idicate borderlie cases which might be argued to ivolve a match. CEx predicate gloss closest matchig dative experiecer meaig verb root(s) argumet argumet available skemmtu etertaimet skemmta léttir relief létta hvatig ecouragemet hvetja ydisauki pleasure auka (+) (+) (+) hressig refreshmet hressa dægradvöl recreatio dvelja (+) skaprau aoyace skapraua vobrigði disappoitmet bregða + + iðurlægig humiliatio iðurlægja álitshekkir reputatio damage hekkja + The theory of polysemy resolutio does ot allow the root of a ou to have meaig properties which are uavailable i the root of a verb if the ou is built o top of that verb. The evidece from allomorphy ad meaig therefore supports the view that the CEx predicate is a ou iterally ad exterally. The followig sectio shows that the dative really is a applied argumet ad that it belogs to the ou phrase rather tha the clause. 3.3 Applicatives are idepedet of verbal morphosytax Dative experiecers/beefactives are i curret theoretical terms itroduced as specifiers of Appl i clausal sytax (Pylkkäe 2000; McGiis 2001; Pylkkäe 2008). A mai distictio is made betwe High Evet Applicatives (29) ad Low Idividual Applicatives (30). (29) High Evet Applicative (idividual-evet; Pylkkäe s High Appl) ApplP DP idividual Appl High VP V... (30) Low Idividual Applicative (idividual-idividual; Pylkkäe s Low Appl) VP V ApplP DP idividual Appl Low DP idividual The High/Low distictio loses some of its termiological appeal i the curret aalysis, ad therefore I will refer to Evet/Idividual applicatives istead (see McGiis 2001). Evet Applicatives relate a idividual to a evet, ad their specifier is typically iterpreted as a beefactive or a experiecer 12

13 by virtue of beig merged ito that positio. Idividual Applicatives relate a idividual to aother idividual ad the specifier of this Appl head is typically iterpreted as a recipiet of the complemet of Appl. Sice we are focusig o the Iceladic CEx costructio, we will oly be cocered with Evet Applicatives here, the type selected by our cause (see 3.4). To accommodate the Iceladic CEx facts, I propose a extesio of the theory, a Root-Selectig Evet Applicative which does ot eed to combie with a verb. 11 (31) Root-Selectig Evet Applicative (idividual- evet: CEx costructio) ApplP DP idividual Appl evet I will ow preset evidece that the dative really is a applied argumet ad that it really is part of the ou phrase ad ot the clause. First, the CEx dative caot be iterpreted as a aget, see (32). It is always a experiecer/beefactive, which suggests that it is of the Spec,Appl type. (32) a. * Þeir sedu vop óviium til eyðileggigar they set weapos eemy.the.dat for destructio Iteded: They set weapos for the (agetive) eemy s destructio b. Dasi var [stelpuum til skemmtuar]. dace.the.om was [girls.the.dat for etertaimet.ge] # The dace had the effect that the girls etertaied somebody. # The girls used the dace to etertai.... It is a potetial aalytical alterative that these are i fact poetic dative possessors (Thráisso 2007:95) rather tha applied argumets. However, the CEx dative cotrasts with dative possessors i that dative possessors alterate with geitives (33) whereas CEx datives alig with Spec,Appl positios i that they do ot participate i case alteratios. Applied datives are resistat to alteratios by sytactic cotext, eve where other datives alterate (Wood 2012:240). (33) a. Er það komið til eyra is it come to ears.ge mér... me.dat It has come to my ears From IcePaHC (Walleberg et al. 2011): 1260.JOMSVIKINGAR.NAR-SAG,.1377 b. Er það komið til eyra mia... is it come to ears.ge my.ge It has come to my ears The example below shows that ulike the dative possessors, the experiecer datives i CEx do ot alterate with geitives. It is ot possible to replace a dative with a geitive i CEx. 12 (34) * Þeir dösuðu stelpaa til skemmtuar * they daced girls.the.ge to etertaimet.ge Iteded: They daced for the girls etertaimet 11 The root still has to be a property of evets, of type s,t. The coceptual idea that little v heads itroduce evets or evetualities (see Maratz 2001, 2007) is i my system derived from a geeral mechaism of polysemy resolutio for roots arbitrated by category heads. Little v (or some flavors thereof) will geerally be compatible with a s,t iterpretatio of the root, but other category heads might too, icludig my causative omializer cause. See 2.2 ad 5.2 for further discussio o polysemy resolutio. 12 It is ot impossible i geeral to have two geitives i a ou phrase. A caoical geitive possessor ca appear with a geitive P complemet as i (33b). The sytax ad sematics of such a getive is differet from CEx datives. For example, it caot appear i a Spec,P positio like a CEx dative ad it is iterpreted as a caoical possessor rather tha a experiecer. 13

14 The dative possessors are furthermore restricted i oly beig compatible with ialieable possessio ad oly followig a prepositioal phrase (Thráisso 2007:95). Note that CEx predicates are sometimes complemets of prepositios but ot always. Based o the evidece ispected so far, I coclude that the girls is a applied dative, ad that etertaimet is a ou both iterally ad exterally. The ext step i the argumetatio is to show that the dative is part of the ou phrase rather tha the clause. We ca use a displacemet test like topicalizatio to show that i (35) the dative is iteral to a structure that is a costituet. (35) [Stelpuum til skemmtuar] dösuðu þeir. [girls.the.dat for etertaimet.ge] daced they They daced for the girls etertaimet The dative is optioally realized to the right of the ou, where it is plausibly i a lower base-geerated positio as illustrated i (36) ad (37). (36) PP DP (girls.the.dat) P for P cause -met ApplP DP (girls.the.dat) Appl exp etertai (37) Þeir dösuðu [(stelpuum) til skemmtuar (stelpuum)]. they daced [(girls.the.dat) for etertaimet.ge (girls.the.dat)] They daced for the girls etertaimet Quatifiers ca be floated i the lower positio as show i (38) ad uder stadard assumptios (Sportiche 1988) such data support the aalysis that there is a movemet relatioship betwee the two positios. (38) Þeir dösuðu [stelpuum til skemmtuar [öllum t(stelpuum)]]. they daced [girls.the.dat for etertaimet.ge [all.dat t(girls.the)]] They daced for all the girls etertaimet Summarizig the fidigs, we ow kow that the girls is a applied dative, ad that etertaimet is a ou, ad that the applied dative is part of the ou phrase. The last part of this sectio discusses the exteral eviromets uder which the CEx costructio is embedded. 3.4 A causative omializer The mai compoets of the curret argumet about applicatives i the ou phrase are (i) to determie that the category head closest to the root is as well as (ii) cofirmig that the dative is a applied argumet which belogs to the ou phrase. However, the realizatio of the same head at the iterface with sematics may hold the key to uderstadig how the CEx costructio eters ito the sytactic eviromets where it is attested. The geeralizatio i (39) is importat to uderstadig the distributio ad it holds for all four varieties of the costructio which were itroduced i 3.1. (39) Obligatory causig evet: If the CEx dative appears with a CEx predicate, a causig evet has to be overtly expressed (as a setece, a DP subject, or a adjoied PP) 14

15 For example, cosider (18) which is repeated as (40). I assume that the (base-geerated) structure is like (41) where a evet deotig subject is itroduced as the causig evet by a appropriate flavor of Pred. For simplicity, I omit from trees (41) ad (42) a fuctioal projectio FP betwee Pred/ whose specifier the dative moves to i (40) ad through i (42). (40) Dasi var stelpuum góð skemmtu. the.dace.om was the.girls.dat good etertaimet.om The dacig etertaied the girls well (41) PredP DP the dace Pred evt{d} was P ap good cause -met P ApplP DP Appl the girls etertai The obligatoriess of the causig evet is highlighted whe we look at a passive-like variat of the CEx costructio i (19), repeated as (42). The by-phrase is most obviously a setece-adjoied PP ad the structure is the like (43) where the dative is the highest argumet ad raises to the sytactic subject positio. It is iterestig that the by-phrase caot be omitted. Adjucts are usually optioal i the sytax ad by-phrases i passives always are. (42) Stelpuum var skemmtu *( af dasium). the.girls.dat was etertaimet.om *( by the.dace) The girls were well etertaied by the dacig (43) PredP PredP PP Pred evt was P by the dace cause -met ApplP DP Appl the girls etertai The proposed aalysis, which will be developed i the dissertatio, is that the causig evet is obligatory for sematic rather tha sytactic reasos. Accordig to this aalysis, the omializer i the CEx costructio is realized at LF with a stadard bievetive causative sematics (Parsos 1990; Pylkkäe 2008) ad Pred evt i the trees above is a piece of meaig that expresses a idetity relatioship betwee a argumet ad the causig evet. (44) a. cause = λp.λe.( e ) P(e ) & CAUSE(e,e ) b. Predevt{D} = Predevt = λp s,t. λe. λe. P(e ) & e = e 15

16 The sematics i (44) ecodes a causative relatioship betwee the evet described by the CEx predicate ad the evet which is amed i Spec,Pred. I follow Wood (2012) i assumig that a head ca vary with respect to whether it requires a exterally merged specifier or ot. Notatioally, a subscript D i a curly bracket idicates that a head demads a DP specifier. The sematics is by hypothesis blid to such sytactic distictios as (44b) shows. Accordig to the aalysis, there is o sytactic reaso for the obligatory by-phrase, but the sematics requires the causig evet to be supplied by whichever available sytactic meas. The causative sematics o the omializer is also hypothesized to accout for the fact that CEx ous are ot compatible with every ou phrase positio ad the ous themselves must be iterpretable as caused evets. For example, a ou like bók book caot be a CEx predicate. The CEx ou is restricted to sytactic positios that are compatible with sematic type s,t. The implemetatio details of this aalysis will be developed i the dissertatio, but the geeral ituitio that I aim to pursue is that the obligatory by-phrase is a key piece of evidece, alog with the fact that the costructio is ituitively iterpreted as a causative. A causative aalysis is further supported by the fact that atoymous adverbial modificatio ca felicitously target the causig evet ad the caused evet: (45) a. Strákarir dösuðu á hættulega hátt gegum eldi stelpuum til the.boys daced i dagerous maer through the.fire the.girls.dat for skemmtuar á hættulausa hátt. etertaimet i safe maer The boys daced i a dagerous maer through the fire for the girls etertaimet i a safe maer. b. # Strákarir dösuðu á hættulega hátt gegum eldi á hættulausa hátt. the.boys daced i dagerous maer through the.fire i safe maer The boys daced i a dagerous maer through the fire i a safe maer. The cotrast shows that while it is ormally ifelicitous to state that the same evet is performed i a dagerous ad a safe maer, the relevat modifiers ca apply separately to the dacig ad the etertaimet i CEx. Therefore, the examples support a bievetive sematics. 3.5 Sectio summary My case study o applicatives i the ou phrase presets a sytactic aalysis which is crucially costraied by observatios about surface pheomea i morphology ad iterpretatio. The fact that we see root-coditioed allomorphy i the CEx ou shows that there is o iteral verbal layer. The closest category head to the root is omial. Iterpretive realizatio also speaks agaist a iteral verbal layer because the CEx ou sometimes expresses a root meaig which caot be expressed by a correspodig verb. The ou-iteral Appl aalysis is supported by further sytactic evidece ad the exteral distributio of the CEx costructio poits towards a bievetive causative aalysis. The overall cosequece is that Evet Applicatives are idepedet of verbal morphosytax, at least i Iceladic. Future directios: I aim to ivestigate i more depth the types of predicates which are compatible with the CEx costructio, focusig o sematic properties which they may have i commo. I aim to carry out a more comprehesive compariso betwee CEx datives ad other ou-associated datives like dative possessors, payig a special attetio to cross-liguistic patters. I aim to work out i more detail the formal implemetatio of the causative sematics ad how it relates to the exteral distributio of the CEx costructio. 16

17 4 Part II: Two defiite articles The secod case study i the dissertatio provides support to the proposal by Schwarz (2009, 2013) that there are two types of defiite articles i atural laguage ad it exteds this view to iclude complex morphological iteractios. 13 The study is based o evidece from Iceladic which has ot previously bee examied i terms of Schwarz s proposal. The fidigs led support to the hypothesis that the two articles are distict morphemes ad that the realizatio of these morphemes ca iteract i distict ways with structural locality coditios. Oe of the two articles has a morphological distributio which is structurally similar to Eglish T(ese) i do-support. I outlie a aalysis of the Iceladic facts i terms of a morphological operatio of local dislocatio uder liear adjacecy. I cotrast the local dislocatio approach with the use of morphological lowerig. I additio to supportig the distictess of two defiite articles, this part of the dissertatio will aim to address the questio of whether local dislocatio ad lowerig (see 2.3) should both be part of our theoretical toolkit or whether the relevat empirical cases should be reduced to local dislocatio. I this sectio of the proposal I will highlight how the curret facts ad aalysis bear o this questio. 4.1 Weak ad strog defiite articles A weak article expresses situatioal uiqueess ad a strog article expresses aaphoricity. The weak article is formalized as follows i the situatio sematics of Schwarz (2009:264). (46) D weak = λs r. λp. ιx. P(x)(s r ) The strog article icludes a additioal aaphoric idex argumet. (47) D strog = λs r. λp. λy. ιx. P(x)(s r ) & x=y I Stadard Germa, the weak article cotracts i certai P(repositio)+D(etermier) cofiguratios but the strog article does ot. Therefore there is a subtle differece i meaig betwee zum ad zu dem i (48). I may usage scearios either type ca be used felicitously, but there are also cotexts which disambiguate the two. (48) a. Has gig zum Haus. Has wet to-the weak house Has wet to the house. b. Has gig zu dem Haus. Has wet to the strog house Has wet to the house. (Schwarz 2009:5) The differece betwee D weak ad D strog i Germa is well disguised because it oly surfaces whe a defiite article with the appropriate set of grammatical features is the sytactic complemet of a prepositio which allows cotractio. May articles are used i eviromets that are ot compatible with P+D cotractio ad the D weak ad D strog are homophoous. It may seem like a distressigly subtle sig of a grammatical distictio that a speaker of Germa sometimes says zum ad sometimes zu dem. However, as show by Schwarz, this surface patter is systematically associated with whether defiiteess is licesed by uiqueess or aaphoricity, ad there is evidece that that the same meaig alteratio is also expressed overtly i other laguages. A parallel distictio betwee articles of uiqueess ad aaphoricity has bee associated with morphological differeces i Ferig Frisia (Ebert 1971 cited i Schwarz 2009), Austro-Bavaria (Simoeko 2013), ad i Aka (Arkoh ad Matthewso 2013). The Aka fidigs are of particular importace sice they show that the pheomeo is ot a property of Germaic (or Ido-Europea). Focusig o Iceladic, I preset a ovel type of argumet for the reality of the two articles. I show that the two types of articles ca iteract with complex structural pheomea i a way that supports 13 The sectio is based o Igaso (2014). 17

18 their distictess i the sytax. As i Germa, the weak/strog distictio is oly overt i a specific type of eviromet i Iceladic. Evaluative adjectives block locality betwee the defiite article ad the ou ad i D weak cotexts, i which defiiteess is licesed by uiqueess rather tha aaphoricity, evaluative adjectives trigger a weak free form article (49a). The free form article is formed by attachig the suffixed article to hi-. The free article has a the-support sytax which is structurally similar to do-support i Eglish clausal sytax (see aalysis i 4.2) ad I will treat hias a support morpheme. I most cases, there is o evaluative adjective i the ou phrase, ad the the defiite article is suffixed (49b-49c) regardless of whether the cotext liceses the weak or the strog article ad the morphology does ot distiguish D weak ad D strog. (49) a. Has er að leika sér með hi fræga bíl. Has is to play refl with the weak famous evaluative car Has is playig with the famous car. b. Has er að leika sér með fræga bíl-i. Has is to play refl with famous restrictive Has is playig with the famous car. c. Has er að leika sér með bíl-i. Has is to play refl with car-the weak/strog Has is playig with the car. car-the weak/strog The alteratio betwee a free form article ad a suffixed article i (49a-49b) has bee associated with stylistic or historical differeces i the literature o the Iceladic ou phrase (see i particular Sigurðsso 1993; Sigurðsso 2006). However, it has also bee oted that the free article is ot compatible with, i Delsig s (1993) terms, deictic referece, or is it compatible with a sigle restrictive adjective (Thráisso 2007). The free article i (49a) is ideed ifelicitous i Moder Iceladic uder a restrictive readig of the adjective ad it souds archaic. However, (49a) souds ormal if the adjective is evaluative ad this is the first metio of the car i the coversatio. The curret sese of evaluative ca be roughly characterized as the presuppositio that the speaker holds the opiio that the evaluatio expressed by the adjective is appropriate i the situatio. This kid of usage depeds o a pragmatic cotext i which a evaluative readig for the adjective is plausible. 14 For example, it is evaluative usage if the cotet of a adjective expresses humor or iroy. Say that Has is a child who ows a toy hovercraft. Has thiks the toy is a car rather tha a hovercraft ad he is always talkig about this car so the ou bíll car deotes a uique referet i a family situatio ivolvig Has. If Has father asks where Has is ad the car has ot bee metioed i the coversatio, his mother ca respod with (49a) i a humorous maer. Example (49b) does ot express the same meaig. 15 Here, the adjective restricts a set of cars ad picks out the famous oe. As i (48), the distictio is rather subtle i (49). However, i cotexts that clearly disambiguate whether defiiteess is licesed by uiqueess or aaphoricity, the weak/strog distictio is clearly draw out i Iceladic as well as Germa. The meaig cotributio of a defiite article is a old liguistic puzzle. O the oe had, it seems obvious that the uiqueess of the ou is importat for the use of the i a expressio like the su. However, this caot be the whole story because it is possible to use a defiite article with ous that describe o-uique thigs. We ca say the book iside a library or a bookstore where the o-uiqueess of the ou is a umistakable property of the immediate situatio. (50) Joh bought a book ad a magazie. The book was expesive. 14 There is a potetially iterestig coectio with the literature o predicates of persoal taste, e.g., i terms of whether the opiio ca be held by someoe other tha the speaker ad how to characterize opiio depedece (see Kölbel 2004; Lasersoh 2005). I leave ivestigatios of such matters for the future. 15 The example with the car ad the hovercraft itroduces a additioal dimesio of complexity by usig a oliteral iterpretatio of a ou. It is ot obvious to me at this poit how to accout for that dimesio. However, the example illustrates that a elaborate cotext may be eeded to support usage of the free article whe a restrictive iterpretatio of the adjective would otherwise be saliet. 18