Gapping of subjec conrol predicaes Luis icene Universiy of California, Sana Cruz koldi@gmail.com April 30, 2008 1 Inroducion Gapping was universally hough be a subcase of ellipsis unil Johnson (1996, 2006) proposed an alernaive analysis in erms of across-he-board movemen. His analysis of an example like (1) ress on he following assumpions: (i) coordinaion happens a he level; (ii) he verb undergoes ATB movemen ou of ; (iii) he subjec of he firs conjunc moves Spec while he subjec of he second conjunc remains in Spec. The whole derivaion is given in (2). (1) Randy drank scoch and Amy [ ] rum. (2) Randy T XP drank and v Amy v scoch rum Johnson s analysis offers some ineresing advanages over deleion approaches gapping. To begin wih, i accouns for he observaion ha negaion and modals necessarily ouscope coordinaion. Moreover, being based on ATB head movemen, i also offers a plausible reason as why gapping is resriced 1
Gapping and conrol Page 2 of 9 coordinaion conexs and resiss embedding. Noneheless, i also raises several non-rivial difficulies. An illuminaing discussion of he laer can be found in Coppock (2001), who concludes ha gapping is bes analyzed as deleion, as radiionally hough. The goal of his squib is offer an addiional argumen agains Johnson s analysis by considering he ineracion of gapping wih subjec conrol predicaes. Our saring poin is example (3): (3) Randy wans a novel and Amy [ ] a play [=...and Amy wans a play] The gapped par in (3) is he whole verbal complex wans. The fac ha he highes verb is a subjec conrol verb helps us pinpoin he lowes possible posiion for he second conjunc subjec Amy: i mus be siing in he Spec of wan, as is exernal argumen. If i were any lower, we would incorrecly predic he possibiliy of a reading in which Randy wans Amy a play. In urn, he lowes possible posiion of he second conjunc subjec also deermines he smalles possible srucure ha can be coordinaed namely, he headed by wan. isually (wih irrelevan projecions omied): (4) and Randy v Amy v wans wans pro T pro T a novel a play In order derive (3) from (4) via ATB movemen, i would be necessary move he whole verbal complex wans. This, however, canno be done direcly, given ha his sring does no form a consiuen. In he upcoming pages, I will consider hree alernaives ha could be invoked achieve his resul. I adop a reducio ad absurdum sraegy, where for each alernaive I will firs idenify he assumpions one would have make in order derive (3) under an ATB movemen analysis, and hen I will show ha hose assumpions make incorrec predicions in oher areas of synax. Thus, he exen ha he reasoning here is correc, we will be led he conclusion ha Johnson s analysis of gapping canno be mainained.
Gapping and conrol Page 3 of 9 2 Non-soluions he problem 2.1 The gapped sring is a consiuen As jus menioned, he main problem wih an ATB analysis of (3) is ha i requires us move a non-consiuen. This difficuly can be circumvened if we choose analyze he sring wans as a consiuen, more or less along he lines ha Koopman and Szabolcsi (2000) have proposed for comparable Hungarian srucures. 1 The derivaion given in (5) below is a raher rough execuion of his idea if we waned be faihful Koopman and Szabolcsi, we would have break i down in a far longer series of movemens. However, he final resul would be idenical in erms of consiuency, so here is no need go in ha level of deail. (5) A remnan movemen analysis of (3) a. Move he objec [a play] [Amy wans ] }{{} b. Move he verb complex [wans } {{ } ] [a play] [Amy ] c. Move he subjec [Amy ] [wans ] [a play] }{{} Afer sep (5c), wans forms a consiuen he exclusion of everyhing else, hence i can undergo ATB movemen in order derive (3). For he sake of argumenaion, le us gran ha all he required movemen operaions can be jusified, and focus on he quesion of wheher i is reasonable call wans a consiuen. The answer is arguably negaive, given ha i his sring does no pass basic consiuency ess. To begin wih, i canno be coordinaed wih a similar sring. Noe ha (6) can be given Righ Node Raising analysis if a play is separaed from he res of he clause by a heavy prosodic break. If (6) is given he regular (non-rnr) innaion ha would correspond he indicaed parse, he resul is ungrammaical. (6) * Randy [waned ] and [ried aend] a play. Similarly, wan(s) canno be lef-dislocaed (7a) or appear in he focus posiion of a clef (7b). 1 Noe ha his is purposefully inaccurae. Koopman and Szabolcsi (2000) argue ha verbal complexes form a consiuen only when he verbs appear in he mirror image of he English order. For srings ha parallel he English order, hey sill assume ha each verb is a separae consiuen (albei one derive by a series of remnan movemens). For he sake of he argumen, I ll ignore his poin.
Gapping and conrol Page 4 of 9 (7) a. * Randy said ha he waned a play, and [wan ], he did a play. b. * I is [wan ] ha Randy does a play. In he hird place, wan(s) canno be a fragmen answer eiher. (8) A: Wha is i wih Randy and he novel? B: * Wan(s). Finally, under his analysis, does no form a consiuen wih a novel. This means ha i should be possible apply ellipsis wihou affecing a novel, and ha i should be impossible elide a novel a leas, he exen ha ellipsis can only affec consiuens. In realiy, he judgemens are he opposie. Noe ha one canno jusify (9) by saying ha he derivaion (5) applies only in he case of gapping. Such an explanaion would only beg he quesion. (9) a. * Randy wans a novel, and Amy wans [ ] a play o b. Randy wans a novel, and Amy wans [ ] o If we allow wans be a consiuen in order derive (3), hen i is no clear why examples (6) hrough (9) should be ungrammaical. In paricular, example (7a) is paricularly myserious, since i illusraes a case of movemen and he same applies (8B), if we adop Merchan s (2004) analysis of fragmen answers in erms of movemen. Given ha he whole poin of analyzing wans as a consiuen was le i undergo movemen, here no reason why i shouldn be able move in hese environmens. In he ligh of hese consideraions, we mus conclude ha his paricular soluion is no plausible. 2.2 Each gapped elemen moves independenly If wans canno be exraced ou of he coordinae srucure as a consiuen, hen i migh be he case ha each of is subconsiuens moves on is own. Under his analysis, (3) would be analyzed as in (10). For ease of parsing, movemens ou of he firs conjunc are represened wih arrows below he example, and movemens ou of he second conjunc wih arrows above i. (10) Randy wans [[ a novel] and [ vp Amy a play]] One can plausibly analyze movemen of he finie verb as movemen a funcional projecion righ ouside v P (see, e.g., Julien 2002). The analysis of he
Gapping and conrol Page 5 of 9 res of he gapped verbs, hough, is no so obvious. Le us assume wihou o much discussion, ha here is a series of funcional heads righ above v P ha provide he required landing sies (cf. he LD/sacking projecions posulaed in Koopman and Szabolcsi 2000). Alernaively, one could also assume muliple adjuncion v P. Eiher way, deriving he correc resul would require an order preservaion consrain (however his is be encoded), so ha he sring of moved verbs parallels he order in is base posiion. Finally, even hough hese movemens arge heads, hey mus be exemped from he very sric localiy condiions on head movemen. These are already quie exraordinary assumpions wihin English synax, bu le us noneheless assume hey are correc. I is emping aack his analysis on he grounds ha i inroduces a grea deal of complicaion in synax. Specifically, since i is possible sack various infiniival predicaes, hings can quickly ge very complicaed. For example, in order represen (11) below in his fashion, i would be necessary draw fifeen separae movemen arrows, each wih is corresponding race. This migh look quie dauning, bu in realiy i is no: as Koopman and Szabolcsi (2000) poin ou, once we accep ha LD projecions (or some equivalen) are available, i is a echnically rivial ask ierae hem derive examples of arbirary size. In any even, he reader will be spared he sigh of he derivaion corresponding (11), and we ll proceed direcly he more subsanial problems raised by his ype of analysis. (11) Randy wans begin ry a novel, and Amy [ ] a play. One imporan quesion concerning (10) is why he non-finie verbs should move. In his respec, i is insrucive consider he ineracion of hese movemens wih ellipsis. In paricular, assume ha Merchan (2007) is correc i claiming ha ellipsis arges a caegory never larger han. If he non-finie verb mus move ouside, hen i follows ha non-finie verbs will never be affeced by ellipsis. In oher words, English would be prediced have a ype of verb-sranding ellipsis (cf. Goldberg 2005) ha would rule in he following example. (12) * Randy wans a novel, and Amy wans [ ] o. In order avoid his resul, one would have say ha he movemens of nonfinie verbs illusraed in (10) only happen in gapping clauses. This, however, leads us in a circular argumen, and we mus conclude ha his is no a possible analysis of (3). 2.3 Small coordinaion plus backward conrol In he previous wo subsecions, we have seen ha i is no possible exrac non-finie verbs ou of he higher, wheher by moving hem as a single consiuen or individually. This paern of exracion is necessary because we have esablished ha he subjec of he second conjunc mus be merged in he Spec of wans. Suppose, however, ha we negae his assumpion and le he
Gapping and conrol Page 6 of 9 second conjunc subjec be merged in he Spec of. This would allow wans be generaed direcly ouside he coordinae srucure. Therefore, gapping could be derived simply by head movemen of a funcional projecion righ ouside v P a much more plausible paern of ATB exracion. The srucure resuling from his analysis is given in (13). For space reasons, I include he srucure only up he level, and assume ha i proceeds in he usual way from here upwards. (13) Randy v wans XP and v Amy v a novel a play As I already menioned in page 2, he main problem wih (13) is ha Amy is no in a posiion where i could receive he θ-role of wans. 2 As far as I can see, here is no way ou of his problem if one assumes ha conrol involves pro, since ha would require Amy be generaed in he Specv P of wans, blocking he ype of coordinaion illusraed in (13). Suppose, however, ha we ake conrol be movemen, as proposed in Hornsein (1999) and subsequen work. In paricular, suppose ha English excepionally allows he phenomenon of backward conrol (where he conroller appears in he expeced posiion of pro), and le us adop he suggesion in Polinsky and Posdam (2002) ha backward conrol can be analyzed as cover movemen in he conex of a movemen heory of conrol. If we gran his much, hen we can recas (13) as (14), where he doed arrow represen he cover movemen of Amy receive is θ-role (order of muliple specifiers is irrelevan here; in order avoid cluering, I have ignored verb movemen). 2 Johnson (2006) includes various examples wih his srucure. However, and very crucially, none of hem feaures a θ-role assigning verb ouside he coordinae srucure, so no problems arise in his respec.
Gapping and conrol Page 7 of 9 (14) Randy v wans XP and v Amy v a novel a play In order for his derivaion work, we mus accep ha a single v head can assign is θ-role wo differen argumens, and ha a single T head can assign case wo differen argumens o. In fac, here is no reason limi his number jus wo: under his analysis, v and T heads mus be able license an arbirary number of independen argumens, as evidenced by he accepabiliy of examples like (15): (15) Randy wans a novel, Amy [ ] a play, Lawrence [ ] a sonne, Rudy [ ] a mah paper, and Enoch [ ] a memoirs book. However, if we accep his, we would also expec find muliple subjecs. Noe ha we are alking here abou rue muliple subjecs, no a single subjec conaining a complex coordinae srucure. In paricular, by analogy wih (14), we would expec (16) have he indicaed reading. (16) * Randy wans Amy a novel. [=Randy wans a novel, and Amy wans a novel o] The derivaion, illusraed in (17) below, would proceed as follows: boh Randy and Amy would be merged as independen exernal argumens of. If we allow one single v head assign more han one exernal θ-role, hen all verbs should in principle be able make use of his possibiliy. Afer ha, Randy raises overly he Spec of wans in order receive a θ-role, and Amy does he same coverly. Finally, he marix T head assigns Case boh Randy and Amy separaely. As above, I only provide he derivaion up he marix level and assume i proceeds normally from here on.
Gapping and conrol Page 8 of 9 (17) Randy v wans Amy v a novel Noe ha we haven invoked any mechanism ha wasn already invoked in (13)/(14), ye (16) is ungrammaical in he indicaed reading. There seems no way capure his asymmery, which leads us he conclusion ha his hird analysis is also unenable. 3 Conclusion We have examined hree differen ways of analyzing gapping of subjec conrol predicaes as ATB movemen, and we have seen ha each analysis runs in very ugh problems. Specifically, we canno exrac wans as a consiuen because i fails all consiuency ess; we canno exrac each of is subconsiuens independenly wihou making very unreasonable assumpions abou English synax; and if we resor a small coordinaion srucure plus backward conrol, hen we creae problems for he heory of argumen licensing. We mus conclude ha, unless a beer analyical opion can be provided, gapping canno be ATB movemen. I find his a raher unforunae conclusion, hough, since one of is consequences is ha all he advanages inheren Johnson s analysis are los as well. 3 3 If we are sric abou i, he only hing ha canno be ATB movemen is gapping wih subjec conrol verbs. I sill could be he case ha, while gapping wih subjec conrol verbs is deleion, regular gapping like (1) is ATB exracion. While his is a logically possible siuaion, i would be exremely srange if he analysis of gapping urned ou spli in his way.
Gapping and conrol Page 9 of 9 References Coppock, Elizabeh. 2001. Gapping: in defense of deleion. In Chicago Linguisics Sociey 37, ed. Andronis, Ball, Elsn, and Neuvel, 133 148. Universiy of Chicago. Goldberg, Lous. 2005. erb sranding ellipsis: a cross-linguisic sudy. Docral disseraion, McGill Universiy. Hornsein, Norber. 1999. Movemen and conrol. Linguisic Inquiry 30:69 96. Johnson, Kyle. 1996. In search of he English middle field. Ms., Universiy of Massachuses, Amhers. Johnson, Kyle. 2006. Gapping is no () ellipsis. Ms., Universiy of Massachuses, Amhers. Julien, Mari. 2002. Synacic heads and word formaion: a sudy of verbal inflecion. Oxford: Oxford Universiy Press. Koopman, Hilda, and Anna Szabolcsi. 2000. erbal complexes. Cambridge, Massachuses: MIT Press. Merchan, Jason. 2004. Fragmens and ellipsis. Linguisics & Philosophy 27:661 738. Merchan, Jason. 2007. An asymmery in voice mismaches in ellipsis and pseudogapping. Linguisic Inquiry 39:169 179. Polinsky, Maria, and Eric Posdam. 2002. Backward conrol. Linguisic Inquiry 33:245 282.