Complement Structures: Outline. Complement Structures and Non-Finite Constructions in HPSG. Problems for Small Clauses. Category Selection

Size: px

Start display at page:

Download "Complement Structures: Outline. Complement Structures and Non-Finite Constructions in HPSG. Problems for Small Clauses. Category Selection"

Emory Garrett
5 years ago
Views:

1 Complement Structures: Outline Complement Structures and Non-Finite Constructions in HPSG Introduction to HPSG 19. Mai 009 Category selection Nonfinite constructions: Raising and rol Passive construction Kordula De Kuty 1 Category Selection Problems for Small Clauses In GB, all of te following bracketed prases are constituents: (1) a. Kim said [(tat) Sandy left]. b. Dana preferred [for Pat to get te job]. c. Leslie wanted [Cris to go]. d. Lee believed [Dominique to ave made a mistake]. e. René tried [pro to win]. f. Tracy proved [te teorem false]. g. Bo considered [Lou a friend].. Gerry expects [tose cildren off te sip]. Tis is te so-called small clause analysis. But tere is a problem... Verbs select for egories witin tose bracketings: () a. I consider [Jon a friend]. b. *I consider [Jon off my sip]. c. I expect [tose cildren off te sip]. d. *I expect [tat man stupid]. But subegorization is assumed to be, i.e., you can t subegorize for someting witin te ting you re subegorizing for. So, eiter it s not or we souldn t use te small clause analysis

2 Questions for analyzing complements Some generalizations Is te selection simply semantic?... Probably not: () a. *I expect tat island off te route. b. I expect tat island to be off te route. Does te need access to te egorial properties of its complements?... Most likely, yes (see above examples) Are all s te same?... Definitely not: expect and consider allow ap, pp, and infinitival complements Wit ap and pp, expect and consider can often ave complements wic are semantically difficult to interpret (e.g., (a)) consider, but not expect, allows np complements () a. I consider/expect tat island to be off te route. b. I consider/*expect tat island off te route. c. I consider/expect tat island to be a good vaion spot. d. I consider/*expect tat island a good vaion spot. Verbs of becoming In fact, lots of s seem to beave pretty differently w.r.t. teir acceptable complements, as in all tese s of becoming: () a. Kim became/grew/got/turned out/ended up/waxed political (AP). b. Kim became/*grew/*got/turned out/ended up/*waxed a success (NP). c. Kim *became/*grew/got/*turned out/*ended up/*waxed sent more and more leaflets (VP-en). d. Kim *became/*grew/*got/*turned out/ended up/*waxed doing all te work (VP-ing). e. Kim *became/grew/got/turned out/*ended up/*waxed to like ancovies (VP-inf). Lexicalization and Localization Complement selection must be igly lexicalized Is not reducible to te semantics of a Generally eld belief tat subegorization is tus, tese s need teir complements to be sisters Verbs of considering sow furter evidence for sisterood of complements Despite similar meanings, none of tese s ave te same paradigm 8

3 Verbs of considering () a. We rate/consider/*regard/*count Kim to be an acceptable candidate (VP-inf). b. We rate/consider/*regard/*count Kim an acceptable candidate (NP). c. We rate/consider/*regard/*count Kim quite acceptable (AP). d. We rate/consider/*regard/count Kim among te most acceptable candidates (PP). e. We rate/*consider/regard/count Kim as an acceptable candidate (PP-as). consider and similar s need access to te prase following Kim Tus, Kim an acceptable candidate doesn t work as a small clause 9 Explaining te facts Tese facts can best be explained if consider takes arguments: te following NP and an XP after tat. () V consider NP Kim VP XP (*as) an acceptable candidate Wit suc a structure, consider can ly constrain te XP But ow is te subject of (to be) an acceptable candidate equated wit te object of consider? 10 Counter-arguments? One counter-argument against suc structures would be if we ad constituency tests sowing tat te small clause analysis was best (8) Tey consider [Kim quite acceptable], and we consider?tat/?it, too. (9) *It was [Kim quite acceptable] tat we considered. In lieu of compelling arguments from constituency tests, eiter we look at teory-internal arguments or take tese subegorization facts seriously.» prase dtrs ed-struc Review: ID Scemata (for Englis):»! loc inv sub -comps-struc dtrs -dtr prase comp-dtrs sign (Head-Subject)»! loc inv sub» (Head-Complement) -comps-struc dtrs -dtr» loc inv + sub (Head-Subject-Complement)» -comps-struc dtrs -dtr 11 1

4 Towards an analysis: Unsaturated Complements Were does it say in tese ID scemata tat every subegorized item must be realized? In Englis, many s and adjectives subegorize for an unsaturated complement. [ In oter s, a complement can be specified as sub NP ], rater tan [ sub ] Te Head-Subject Scema allows for tis. And tis will give consider access to te lower subject, as well as its own subject. Te empirical callenge of non-finite constructions In non-finite constructions, te subject of te embedded is not expressed as a ly realized dependent. Problem 1: Wat is interpreted to be te subject of te non-finite? (10) a. Jon tried to dance. (subject) b. Jon promised Peter to dance. (subject) c. Jon persuaded Peter to dance. (object) 1 1 Problem : Wy do s selecting non-finite complements differ w.r.t. wat kind of rollers can occur? (11) a. Jon tried to dance. (subject) b. Jon appeared to dance. (subject) (1) a. Jon persuaded Peter to dance. (object) b. Jon expects Peter to dance. (object) (1) a. * It tried to rain. (subject) b. It appeared to rain. (subject) (1) a. * Jon persuaded it to rain. (object) b. Jon expects it to rain. (object) Classifying non-finite complements Verbs selecting non-finite complements can be classified according to teir orientation (subject, direct or indirect object): Wat is interpreted to be te subject of te non-finite complement? te nature of te relationsip of te embedding to te roller: Is te roller an argument of te embedding? 1 1

5 Empirical basis of classifiion I. Orientation: Determined by interpretation of embedded infinitive. II. Relationsip: embedding roller Raising s only mediate te requirement of te complement: 1. If te embedded requires a non-referential (= dummy) subject, so does te raising : Te same olds wen a dummy subject is permitted: (18) a. Tere is a dragon in te wood. b. Hobbs is a dragon in te wood. (19) a. Tere seems to be a dragon in te wood. b. Hobbs seems to be a dragon in te wood. (0) a. * Tere wants to be a dragon in te wood. b. Hobbs wants to be a dragon in te wood. (1) a. It rains. b. * God rains. (1) a. It seems to rain. b. * God seems to rain. (1) a. * It wants to rain. b. * God wants to rain If te embedded permits a clausal subject, so does te raising : (1) a. Tat Sara dances fascinates Jon. b. Sara fascinates Jon. () a. Tat Sara dances seems to fascinate Jon. b. Sara seems to fascinate Jon. () a. * Tat Sara dances wants to fascinate Jon. b. Sara wants to fascinate Jon.. If te embedded as a subject wit an idiomatic interpretation, so does it wen selected by a raising : () Te is out of te bag. () Te seems to be out of te bag. () % Te wants to be out of te bag.. In languages were subjectless constructions exist, raising s can embed suc subjectless complements: () a. Dort sceint getanzt zu werden. tere seems danced to be People seem to dance over tere. b. In sceint zu frieren. im seems to freeze He seems to freeze. (8) a. * Dort versuct getanzt zu werden. tere tries danced to be b. * In im versuct tries zu to frieren. freeze 19 0

6 . passivization of te non-finite complement results in a paraprase (9) a. Jon seems to read a book. b. Te book seems to be read by Jon. (0) a. Jon wants to read a book. b. % Te book wants to be read by Jon. 1 Subject oriented raising s pon <seem> [ vform bse ] sub 1, VP [ inf, sub 1 ] : [ seem soa-arg ] Subject oriented equi s pon <try> [ vform bse ] sub NP 1, VP [ inf, sub NP 1 ] : try tryer 1ref soa-arg Object oriented raising s pon <believe> [ vform bse ] sub NP 1,, VP [ inf, sub ] : believe believer 1ref soa-arg

7 Object oriented equi s pon <persuade> [ ] vform bse [ sub NP 1, NP, VP inf, sub ] NP : persuade persuader 1ref persuadee ref soa-arg Raising vs. Equi s Equi (rol) s assign a semantic role to all arguments; raising s do not Equi s co-index teir subject (object) wit an argument s unexpressed subject; raising s completely structure-sare tese elements Equi s require a referential subject (object), wile raising s do not Capturing te generalization beind raising Raising as a widespread penomenon Complements of oter egories Raising Principle (Pollard and Sag 199, p. 10) Let E be a lexical entry wose sub list L ains an element X not specified as expletive. Ten X is lexically assigned no semantic role in te ent of E if and only if L also ains a (nonsubject) Y [sub < X >]. Not only VP but also AP complements are possible complements of raising s like seem or appears: (1) Jon seems/appears intelligent. () Jon seems/appears to be intelligent. 8

8 Raising as a widespread penomenon More on AP complements Interesting classes of raising s I Suc AP complements beave parallel to te VP complements of raising s, e.g., wit respect to clausal subjects: () a. Tat e came to er wedding is tasteless. b. Jon is tasteless. () a. Tat e came to er wedding seems to be tasteless. b. Jon seems to be tasteless. () a. Tat e came to er wedding seems tasteless. b. Jon seems tasteless. () a. * Tat e came to er wedding wants to be tasteless. b. Jon wants to be tasteless. Copula be + prediive XP: () Jon is ugly. (AP complement) e.g., non-referential subjects of AP complements are possible: (8) It is likely tat Jon sings tonigt. Certain modal s + VP[bse]: (9) Jon may/must/sall run. (0) It may/must/sall rain. 9 0 Interesting classes of raising s II Te infinitival marker to Tense auxiliaries: (1) is + gerund a. Jon is running ome. b. It is raining. () will + base form: a. Jon will run ome. b. It will rain. () ave + past-participle: a. Jon as run ome. b. It as rained. () a. Jon will laug / *to laug / *lauged / *lauging. b. Jon as *laug / *to laug / lauged / *lauging. c. Jon is *laug / *to laug / *lauged / lauging. d. Jon seems *laug / to laug / *lauged / *lauging. Wat is te status of te infinitival marker to? Is it part of te morpology of te infinitive, or a separate syntactic element? 1

9 Arguing for status of to as separate syntactic element Wat kind of an element is to? A raising! In coordination exts to can select a prase: () Jon seems [to [come and go]] as e pleases. (8) Jon seems to run. (9) It seems to rain. A sample raising analysis: Te alternative analysis () Jon seems [[to come] and [go]] as e pleases. is implausible since te form of te coordination is neiter only determined by te first conjunct nor always to-infinitive if one is present: () * Jon seems [[go] and [to come]] c it It seems to rain seems to rain c seems to rain c to rain Te subtree to rain prase p <to rain> " s 8l subc p <to> " vform inf s l D subc, 1VP bse, subc i E : c p <rain> " s 1 l vform bse subc NP it rain c p <it> " s l noun subc ˆindex it Analysis of: It seems to rain p <seems> " 9 vform fin D s l subc, 8VP inf, subc " seem 10 soa-arg prase p <It seems to rain> " s l 9 subc 10 prase p <seems to rain> " s l 9 subc 10 i : prase p <to rain> " E s 8 l vform inf subc NP it rain c

10 Analysis of: *It wants to rain p <wants> s l 9 " vform fin subc D NP 11, 8VP inf, subc D NP 11 Ei : E con 10 want wanter 11ref soa-arg prase p <to rain> s 8 l " vform inf subc D NP it E con rain c prase p <wants to rain> s l " 9 subc con 10 Te Passive: subject-to-subject raising (0) a. Jon as stolen te ring. b. Te ring was stolen. pon <as>» vform fin sub D 1, VP past-participle, sub 1 i : E» perfective soa-arg pon <was>» vform fin sub D 1, VP passive-participle, sub 1 i : E 8 Past and Passive Participles pon <stolen>» vform past-participle sub DNP 1, NP E steal stealer 1ref stolen ref pon <stolen>» vform passive-participle sub DNP E D PP[by] 1 E steal stealer 1ref stolen ref 9 Past and Passive Participles A lexical rule to express te generalization loc " vform past-participle sub D NP 1, NP E loc " vform passive-participle sub DNP E D PP[by] 1 E 0

Handout 3 Verb Phrases: Types of modifier. Modifier Maximality Principle Non-head constituents are maximal projections, i.e., phrases (XPs).

Handout 3 Verb Phrases: Types of modifier Modifier Maximality Principle Non-head constituents are maximal projections, i.e., phrases (XPs). Compare buy and put: (1) a. John will buy the book on Tuesday.