Long Dstance Wh-movement n Seereer: Implcatons for Intermedate Movement Nco Baer U Berkeley nbbaer@berkeley.edu PL 38 March 29, 2014 1 Introducton Queston: What motvates ntermedate movement n a successve-cyclc long dstance dependency? (1) [ P What do you thnk [ P <> [ John bought <>? ]] 2 1 Two broad styles of analyss are present n the lterature: Termnal movement and ntermedate movement are derved n the same way (homsky 1995; Mcloskey 2002; Abels 2012; van Urk and Rchards 2013). Intermedate movement s trggered va a dfferent mechansm than termnal movement (Heck and Müller 2000; homsky 2000; Boškovć 2007) Ths talk wll examne data from long dstance wh-dependences n Seereer, a language n whch successve cyclc movement leaves two knds of marks on the clause: Each verb along the path of the dependency bears specal morphology. A pronoun reflectng the φ-features of the moved wh-phrase appears at the edge of each embedded clause. (2) xar foog-o Yande a-lay-u thnk-2sg.ext 3SG Yande 3-say-EXT 3SG What do you thnk Yande sad saw? Seereer resembles languages that show morphologcal reflexes of successve cyclc movement, such as Irsh (Mc- loskey 2002), Knande (Schneder-Zoga 2009), or hamorro (hung 1994) and languages that show repetton n Ā-chans, such as German (Felser 2004), Dutch (Barbers et al. 2009), or Passamaquoddy (Bruenng 2006). I show that these pronouns are actually overtly spelled out copes of the moved wh-phrase and that morphology on the verb spells out the feature that derves Ā-movement. I argue that these two phenomena provde evdence that ntermedate movement steps are just lke termnal movement. To capture ths, I propose the followng: Both termnal movement and ntermedate movement are feature-drven (homsky 1995; Mcloskey 2002; Abels 2012). Each round of movement to the edge of a P counts as ts own Ā-chan; ths s allows n multple copy spell-out. I thank Lne Mkkelsen and Peter Jenks for nsghtful comments, gudance, and dscusson on ths project, as well as my consultant El Hadj Malck Loum for sharng hs language wth me. All data n ths paper were gathered durng the 2012-2013 U Berkeley Feld Methods class and subsequent follow up work wth John Merrll n 2013-2014. Abbrevatons: AGR = agreement; AUX = auxlary; L = class; DET = determner; DV = default vowel; EXT = extracton suffx; FUT = future; IMPF = mperfectve; INF = nfntve; LEP = left edge pronoun; OBJ = object; PL = plural; PST = past; REFL = reflexve; REL = relatve; SBJ = subject; SG = sngular; 1 = frst person; 2 = second person; 3 = thrd person. 1
Nco Baer 2 Verbal Morphology and Ā-movement n Seereer In Seereer, verbal morphology s senstve as to whether or not Ā-movement has occurred. Ths senstvty manfests tself n whch form of fnal suffx a verb takes. Verbs n declaratve clauses take the fnal suffx -a, as n (3). Verbs n wh-questons take the fnal suffx -u, as n (4): (3) Declaratve a-jaw- a 3-cook-DV maalo rce fe DET (4) Wh-queston xar a-jaw- u 3-cook-EXT? cooked rce. What dd cook? Wh-questons also nvolve frontng of the wh-phrase. The fnal suffx -u also appears n focus clauses, (5a), and relatve clauses, (5b) 1 : (5) a. maalo FO Mataar a-jaw- u rce Mataar 3-cook-EXT Mataar cooked RIE FO. b. maalo fe [ P Mataar a-c - uu -n-a rce DET Mataar 3-cook-EXT-3OBJ-REL the rce that Mataar gave hm.. Focus lause Relatve lause Generalzaton: The suffx -u appears n cases of Ā-movement. I wll refer to ths suffx as extracton morphology (glossed EXT). Because of the suffx s senstve to Ā-movement, I propose that the fnal suffxes are located n the -layer, specfcally Rzz s (1997) Fn 0. The verb undergoes head movement to Fn 0 durng the dervaton: (6) Fnal Suffxes Occupy Fn 0 FnP V + Fn 0 -a DV / -u EXT TP An argument for V-to- movement comes from the fact that a fnal suffx s blocked when there s a hgher auxlary n the clause: (7) a. a-xe 3-AUX ret-aa-(*a) go-impf-dv s gong to. b. xan-a FUT-3 ret-(*a) go-dv wll go to. Progressve Aux Future Aux I assume that the hghest verbal element moves to Fn 0. When the the lexcal verb moves, the fnal suffx surfaces. Auxlares have specal paradgms. 1 In (5b), -u s lengthened due to a regular morphophonologcal rule trggered by the relatve suffx -(n)a. 2
Nco Baer The morphologcal expresson of subject-verb agreement s also senstve to extracton n Seereer 2 : Fnal suffxes have specal forms for 1SG and 2SG subjects. Subject extracton surpresses subject φ-agreement ( Ant-Agreement ; Ouhalla 1993) 3. Declaratve Extracton (8) SINGULAR PLURAL 1 V-aam -V M -a 2 V-aa nu-v M -a 3 a-v-a a-v M -a (9) SINGULAR PLURAL 1 V-um -V M -u 2 V-o nu-v M -u 3 a-v-u a-v M -u AAE V-u V M -u To capture the fact that the fnal suffxes n nteract wth φ-agreement, I propose the followng: Fn 0 s the -level phase head n Seereer and t carres both an unvalued OP-feature whch derves Ā-movement (Mcloskey 2002) and unvalued φ-features whch derve subject agreement and subject movement. Snce Fn 0 has two movement nducng probes, t can project to specfers. Examples (10) and (11) show how ths works for non-subject and subject wh-questons, respectvely. (10) Non-subject wh-movement a. Subject DP moves to nner Spec-FnP to satsfy [uφ]. b. Wh-phrase moves to outer Spec-FnP to satsfy [uop]. c. Assume no tuckng n of an Ā-operator nsde an A-poston. d. FnP DP WH [ OP] DP [ SBJ φ] V [ + Fn ] 0 uφ uop TP... <DP WH>... (11) Subject wh-movement a. Subject DP satsfes both [uφ] and [uop] on Fn 0. b. Only one Spec-FnP. c. FnP DP WH.SBJ [ φ OP ] V [ + Fn ] 0 uφ uop TP... <DP WH.SBJ >... When uop on Fn 0 s valued, the extracton paradgm s used. When uop remans unvalued, the declaratve paradgm s used. I follow Premnger (2011) n postng that the OP-probe s allowed to fal, and therefore ths feature can always be present on the P phase head. Upshot: Wh-movement to Spec-FnP s feature-drven. Whether or not the feature dervng movement s valued has morphologcal consequences. 2 In (8) and (9), V M represents a verb stem that has undergone plural ntal consonant mutaton. 3 For reasons of space and tme, I wll not propose an analyss of agreement here. 3
Nco Baer 3 Long Dstance Wh-Movement n Seereer Long dstance wh-movement n Seereer s characterzed by two mportant propertes: Each verb along the path of the dependency bears extracton morphology. A pronoun reflectng the φ-features of the moved wh-phrase appears at the edge of each embedded clause. (12) (13) xar foog-o thnk-2sg.ext 3SG What do you thnk bought? a-jk-u 3-buy-EXT ann foog-o den Yande a-lay-u who.pl thnk-2sg.ext 3PL Yande 3-say-EXT Who all do you thnk Yande sad bsaw? den 3PL I wll refer to the pronoun present n these clauses as a left edge pronoun (abbrevated LEP). Left edge pronouns occur wth all wh-phrases n Seereer n cases of long dstance extracton. I wll focus on cases of argument extracton n ths talk. Both extracton morphology and the LEP are oblgatory. Omsson of ether results n ungrammatcalty: (14) a. * xar foog-o thnk-2sg.ext Ø a-jk-u 3-buy-EXT Intended: What do you thnk bought? No pronoun b. * xar foog-o a-jk- a thnk-2sg.ext 3SG 3-buy-DV Intended What do you thnk bought? Suffx -a The pattern s schematzed n (15): (15) The Pattern of Long Dstance Wh-movement [ P WH... V-u... [ P LEP... V-u... ]] The poston of LEPs n embedded clauses s fxed. As seen n (16), they must always surface n the left perphery of the embedded clause, followng an overt complementzer: (16) a. xar xalaat-o thnk-2sg.ext LEP What do you thnk saw? b. * xar xalaat-o [ P yee thnk-2sg.ext LEP c. * xar xalaat-o thnk-2sg.ext I argue that the LEP occupes the same structural poston as a wh-phrase, Spec-FnP. The complementzer (y)ee occupes a hgher head, namely Force 0 : LEP 4
Nco Baer (17) Poston of LEP ForceP Force 0 yee FnP LEP (SUBJ) V + -u TP When the LEP corresponds to an object, t occupes the outer Spec-FnP, as per (10). When t corresponds to a subject, t occupes the only Spec-FnP, as per (11). The presence of oblgatory extracton morphology n each clause suggests that the LEP reaches ts surface poston va movement. Movement of the LEP values the uop-features on Fn 0 and ths results n the verb n that clause havng extracton morphology. Ths s shown for a non-subject LEP n (18): (18) Non-subject LEP movement ForceP Force 0 (y)ee LEP FnP SBJ V + Fn 0 -u TP... <LEP>... onsequence: Ā-movement n the matrx clasue and n the embedded clause s feature-drven. I propose that all P phase heads carry features that derve Ā-movement, but that these probes are allowed to fal (Premnger 2011; Abels 2012). Ths means they can be present even when there s no long dstance dependency. The fact that movement occurs n embedded clauses does not tell us anythng about the relatonshp between the matrx wh-phrase and the LEP. Below, I show they are related drectly through movement. 4 Left Edge Pronouns as opes Queston: What s the relatoshp between the matrx wh-phrase and the LEP? Opton 1: The wh-phrase and the pronoun are lnked drectly va movement, lke some analyses of wh-copyng by Felser (2004); Barbers et al. (2009); Schppers (2012). (19) The Pattern of Long Dstance Wh-movement [ P WH... V-u... [ P LEP... V-u...... <WH > ]] Opton 2: The wh-phrase and the pronoun are base generated separately and related ndrectly, such as some analyses of wh-scope markng by Dayal (1994); Herburger (1994); Legate (2011). (20) [ P WH... V-u... [ <WH > [ P LEP k... V-u...... <LEP k > ]] 5
Nco Baer I argue for Opton 1: Seereer left edge pronouns are overtly spelled out copes of the moved wh-phrase. Arguments for ths analyss come from: Islands Reconstructon for reflexve bndng 4.1 Argument 1: Islands Polar questons, relatve clauses, and adjunct clauses all behave as slands for Ā-movement n Seereer. The presence of a left edge pronoun nsde an sland does not allevate a volaton ncurred from formng a whdependency wth a gap nsde that sland. Extracton from a embedded polar queston wth the complementzer ndax s llct, whether or not a LEP s present. (21) Embedded Polar Queston a. * xar and-o know-2sg..ext [ P ndax INT Am LEP Am Intended: What do you know whether Am saw? and-o [ P ndax Ø Am know-2sg.ext INT Am Intended: What do you know whether Am saw? b. * xar Extracton from nsde a relatve clause s also llct: (22) Relatve lause Island (PN) a. * an who ga -o see-2sg.ext [ DP maalo rce fe DET [ P jaw-na LEP cook- REL Intended: Who dd you see the rce that cooked? b. * an ga -o [ DP maalo fe [ P Ø jaw-na who see-2sg.ext rce DET cook- REL Intended: Who dd you see the rce that cooked? And the same s true of adjunct slands: (23) Adjunct Islands a. * xar ret-o go-2sg.ext [ P balaa.before LEP a-jk-u 3-buy-EXT Intended: What dd you go to before bought? ret-o [ P balaa Ø go-2sg.ext.before b. * xar a-jk-u 3-buy-EXT Intended What dd you go to before bought? Because slands are standardly taken to be domans that block movement, these facts can be taken to support the dea that left edge pronouns and the matrx wh-phrase are connected by movement. 6
Nco Baer If the matrx wh-phrase and the pronoun were base generated separately, we would not expect these effects to occur. Ths s because an sland boundary would not be crossed by movement n any of the examples n (21), (22), and (23) In fact, under the base generated account, we mght even expect the presence of a LEP to allevate sland volatons, as base generated resumptve pronouns often do (Mcloskey 2002; Schneder-Zoga 2009). 4.2 Argument 2: Reconstructon Left edge pronouns behave as f they are copes of the moved wh-phrase for purposes of reflexve bndng. Frst, notce that n (24), the reflexve complement of the noun foto photo, pcture may be bound by the embedded subject Yande but not the matrx 3PL subject (here ndcated only by agreement): (24) a j -nqalaat-a 3-thnk.PL-DV Yande Yande a-ga -a 3-see-DV foto photo They thnk Yande saw a pcture of herself/*themselves. xoox um /*den j REFL 3SG/3PL However, when the wh-phrase s extracted, the bndng possbltes for the reflexve expand: bndng by the matrx subject s now lct. (25) [foto photo xoox um /den j REFL 3SG/3PL num] k whch a j -nqalaat-u 3-thnk.PL-FO ten k 3SG Whch pcture of herself/themselves do they thnk Yande saw? Yande Yande 3-see-DV k Ths expanson of bndng posbltes s due to reconstructon to the poston of the LEP. Followng Followng Lee- Schoenfeld (2008), I and take the bndng doman of a reflexve to be the phase. Thus, the poston of the LEP at the edge of the embedded P can be bound by the matrx subject. BINDING (26) [whch photo of self] they know [ P [whch photo of self] [ Yande sees [whch photo of self] ] REONSTRUTION Takng reconstructon to straghtforwardly follow from movement (Sportche 2007), ths ponts to the LEP beng an overt copy of the matrx wh-phrase. Notce that the reflexve can stll be bound by the embedded subject. In ths, we have support for movement connectng all three postons. If the matrx wh-phrase and the LEP are generated seperately, reconstructon to the most embedded poston would not be expected. oncluson: Left edge pronouns are the overtly spelled out copes of the moved wh-phrase. 7
Nco Baer 5 Multple opy Spell-out and Ā-chans Problem: Spellng out multple copes of the same tem should be problematc for lnearzaton (Kayne 1994; Nunes 2004) 4. Why and how do multple copes surface? Observaton: In Seereer, copes surface only n Spec-FnP, the poston that Ā-movement targets n termnal movement. opes cannot surface n ther base generated poston: (27) * xar xalaat-o thnk-2sg.ext Intended: What do you thnk saw? I post that copes surface n Spec-FnP because movement to that poston s feature-drven. Here, I assume that the movement dervng features are uφ and uop. More precsely, I argue that movement to such postons defnes a copy there as the head of a chan. I propose the han Head Prncple to formalze ths dea: LEP (28) The han Head Prncple: A copy n a chan s dentfed as the head of that chan f t s nvolved n the valuaton of [uφ] or [uop]. Thus, we have two types of chan heads: Those that value uφ (= A-chan head) Those that value uop (= Ā-chan head) The prncple n (28) has a radcal consequence: Successve cyclc Ā-movement generates multple Ā-chans, each termnatng n Spec-P. Gven ths consequence, termnal movement and ntermedate movement are dentcal: both are feature-drven and both create a sngle chan. Thus, successve cyclc movement s the result of multple chans beng nterleaved. onsder ths means for a long dstance dependency wth two levels of embeddng, as n (29a): (29) a. xar foog-o Yande thnk-2sg.ext 3SG Yande What do you thnk Yande sad saw? a-lay-u 3-say-EXT 3SG b. [ ForceP xar... V+Fn 0 [ ForceP... [ FnP... V+Fn 0 [ ForceP... [ FnP... V+Fn 0 <xar >? ]] han 3 han 2 han 1 Ths s not just a termnologcal dfference! It can be leveraged to allow multple copy spell-out. Because there are multple chans n (29b), the structure can be lnearzed wth multple copes. Each chan head can be targeted for spell-out. In other words: The heads of separate chans are lnearzed as separate tems. 4 A smlar problem arses n the system of yclc Lnearzaton developed n Fox and Pesetsky (2005). 8
Nco Baer Ths dea s ndependently motvated, chans nvolvng dfferent arguments must be treated separately: (30) [ P What k dd [ TP John j [ vp <John j > buy < >?]]] New Problem!: Why do the heads of ntermedate chans surface n some languages (e.g., Seereer, German) but not n others (e.g., Englsh). Notce that there s an added layer of complexty. Take an Englsh example: D B A (31) [ P What do you thnk [ P [ Mary sad [ P [ John bought? ]] HAIN 3 HAIN 2 HAIN 1 Intermedate chan heads dffer from the matrx copy and the copy at the foot of the dependency n an mportant way: they are both the head of one chan, and the tal of another: opy B s smultaneously the head of han 1 and the tal of han 2. opy s smultaneously the head of han 2 and the tal of han 3. I post that ths results n a representatonal ambguty at the pont of spell-out. The nterface does not know whether to treat ntermedate copes as the head of one chan or the tal of another. I propose that the nterface does not tolerate such ambguty. However, languages dffer n the way they resolve ths ssue, dctated by a spell-out paramter: (32) The Heads/Tals Parameter An tem must be unambguously dentfed as the head or tal of a chan at the pont of spell-out. When an tem s ambguous as to head/tal status, languages resolve ths dfference n one of two ways:. That tem s treated as the tal of a chan. (non-wh-copyng languages). That tem s treated as the head of a chan. (wh-copyng languages) Englsh has settng (); Seereer has settng (). Ths results n a structure lke (31) beng spelled out dfferently: (33) a. In both languages, the hghest copy (D) wll always be spelled out, as t wll always be a chan head. b. In both languages, the lowest copy (A) wll never be spelled out, as t always be a chan tal. c. In Seereer, ntermedate copes (B, ) wll always be spelled out because they are treated as heads. d. In Englsh, ntermedate copes (B, ) wll never be spelled out because they are treated as tals. The Heads/Tal parameter predcts that we wll fnd no language n whch some ntermedate postons are spelled out and others are not. Ths s borne out n the lterature (Schppers 2012) The Heads/Tals parameter does not dctate how a gven copy wll be spelled out. It s only functon s to resolve the representatonal ambguty of ntermedate copes created by successve cyclc movement Dfferent languages can therefore be expected to dffer n the way that multple copes are realzed on the surface. For example, n wh-copyng languages lke German, ntermedate copes must be spelled out dentcally n form to the matrx wh-phrase: (34) Wen who glaubst thnk du, you wen who se she getroffen met hat? has Who do you thnk she met? (German, Felser 2004:544) Under the vew advocated here, such dfferences must must be analyzed as arsng from other ndependent propertes of those languages morphology. Ths s n lne wth the Mnmalst deal of locatng varaton n the nventory of lexcal tems and at the nterfaces (n ths case, PF), nstead of n the operaton of the narrow syntax proper. 9
Nco Baer 6 oncluson In ths talk, I have shown that n Seereer long dstance wh-dependences, ntermedate movement looks just lke termnal movement: Both termnal and ntermedate movement trgger extracton morphology on the verb. Both termnal and ntermedate movement result n the spellng out of a copy of the moved wh-phrase. I developed an analyss n whch both termnal movement and ntermedate movement are drven by a feature on the P-level phase head, n Seereer s case Fn 0 (homsky 1995; Mcloskey 2002; Premnger 2011; Abels 2012; van Urk and Rchards 2013). I further proposed that feature-drven movement defnes the end of chans. Therefoe, each step of movement to a P phase edge counts as ts own Ā-chan, and that ths property s results n multple copy spell-out n Seereer. 10
Nco Baer References Abels, Klaus. 2012. Phases: An essay on cyclcty n syntax. De Gruyter. Barbers, Sjef, Olaf Koeneman, and Marka Lekakou. 2009. Syntactc doublng and the structure of wh-chans. Journal of Lngustcs 45:1 46. Boškovć, Željko. 2007. On the localty and motvaton of Move and Agree: An even more mnmal theory. Lngustc Inqury 38:589 644. Bruenng, Benjamn. 2006. Dfferences between the wh-scope-markng and wh-copy constructons n Passamaquoddy. Lngustc Inqury 37:25 49. homsky, Noam. 1995. The mnmalst program. ambrdge, Massachusetts: MIT Press. homsky, Noam. 2000. Mnmalst nqures: The framework. In Step by step: Essays on mnmalst syntax n honor of Howard Lasnk, ed. Roger Martn, Davd Mchaels, and Juan Uragereka, 89 156. MIT Press. hung, Sandra. 1994. Wh-agreement and referentalty n hamorro. Lngustc Inqury 25:1 44. Dayal, Veneeta Srvastav. 1994. Scope markng as ndrect wh-dependency. Natural Language Semantcs 2:137 170. Felser, lauda. 2004. Wh-copyng, phases, and successve cyclcty. Lngua 114:543 574. Fox, Danny, and Davd Pesetsky. 2005. yclc lnearzaton of syntactc structure. Theoretcal Lngustcs 31:1 46. Heck, Faban, and Gereon Müller. 2000. Successve cyclcty, long-dstance superorty, and local optmzaton. In WFL 19. Herburger, Elena. 1994. A semantc dfference between full and partal wh-movement n german. Paper presented at the LSA Annual Meetng, Boston. Kayne, Rchard S. 1994. The antsymmetry of syntax. ambrdge, Massachusetts: MIT Press. Lee-Schoenfeld, Vera. 2008. Bndng, phases, and localty. Syntax 11:281 298. Legate, Jule Anne. 2011. Warlpr wh-scope markng. Syntax 14:97 121. Mcloskey, James. 2002. Resumpton, successve cyclcty, and the localty of operatons. In Dervaton and explanaton n the mnmalst program, ed. Samuel Davd Epsten and T. Danel Seely, 184 226. Malden, Massachusetts: Blackwell Publshng. Nunes, Jaro. 2004. Lnearzaton of chans and sdeward movement. Lngustc Inqury Monographs. ambrdge, Massachusetts: MIT Press. Ouhalla, Jamal. 1993. Subject-extracton, negaton and the ant-agreement effect. Natural Language and Lngustc Theory 11:477 518. Premnger, Omer. 2011. Agreement as a fallble operaton. Doctoral Dssertaton, MIT. Rzz, Lug. 1997. The fne structure of the left perphery. In Elements of grammar, ed. Llane Haegeman, 281 337. Dordrecht: Kluwer Academc Publshers. Schppers, Ankelen. 2012. Medal copy spell-out n long-dstance wh-dependences. In Proceedngs of onsole XVII. Schneder-Zoga, Patrca. 2009. Whagreement and bounded unbounded movement, volume Mergng Features. Oxford Unversty Press. Sportche, Domnque. 2007. Reconstructon, bndng and scope. In The blackwell companon to syntax, ed. Martn Everaert and Henk van Remsdjk. Blackwell Publshng, Malden, MA, USA. van Urk, oppe, and Norvn Rchards. 2013. Two components of long-dstance extracton: Successve cyclcty n dnka. MS: MIT. 11