Complement Structures: Outline Complement Structures and Non-Finite Constructions in HPSG • Category selection • Nonfinite constructions: Raising and control • Passive construction Introduction to HPSG 19. Mai 2009 Kordula De Kuthy 1 2 Category Selection Problems for Small Clauses This is the so-called small clause analysis. But there is a problem ... In GB, all of the following bracketed phrases are constituents: Verbs select for categories within those bracketings: (1) a. Kim said [(that) Sandy left]. (2) a. I consider [John a friend ]. b. Dana preferred [for Pat to get the job]. b. *I consider [John off my ship ]. c. Leslie wanted [Chris to go]. c. I expect [those children off the ship ]. d. Lee believed [Dominique to have made a mistake]. d. *I expect [that man stupid ]. e. Ren´ e tried [ pro to win]. f. Tracy proved [the theorem false]. But subcategorization is assumed to be local , i.e., you can’t subcategorize g. Bo considered [Lou a friend]. for something within the thing you’re subcategorizing for. h. Gerry expects [those children off the ship]. • So, either it’s not local or we shouldn’t use the small clause analysis 3 4
Questions for analyzing complements Some generalizations • Is the selection simply semantic? ... Probably not: (3) a. *I expect that island off the route. • expect and consider allow ap , pp , and infinitival complements b. I expect that island to be off the route. • With ap and pp , expect and consider can often have complements which • Does the verb need access to the categorial properties of its complements? are semantically difficult to interpret (e.g., (3a)) ... Most likely, yes (see above examples) • consider , but not expect , allows np complements • Are all verbs the same? ... Definitely not: (4) a. I consider/expect that island to be off the route. b. I consider/*expect that island off the route. c. I consider/expect that island to be a good vacation spot. d. I consider/*expect that island a good vacation spot. 5 6 Verbs of becoming Lexicalization and Localization In fact, lots of verbs seem to behave pretty differently w.r.t. their acceptable complements, as in all these verbs of becoming: • Complement selection must be highly lexicalized (5) a. Kim became/grew/got/turned out/ended up/waxed political (AP). – Is not reducible to the semantics of a word b. Kim became/*grew/*got/turned out/ended up/*waxed a success (NP). • Generally held belief that subcategorization is local → thus, these verbs c. Kim *became/*grew/got/*turned out/*ended up/*waxed sent need their complements to be sisters more and more leaflets (VP-en). – Verbs of considering show further evidence for sisterhood of comple- d. Kim *became/*grew/*got/*turned out/ended up/*waxed doing ments all the work (VP-ing). e. Kim *became/grew/got/turned out/*ended up/*waxed to like anchovies (VP-inf). Despite similar meanings, none of these verbs have the same paradigm 7 8
Verbs of considering Explaining the facts These facts can best be explained if consider takes 2 arguments: the following NP and an XP after that. (6) a. We rate/consider/*regard/*count Kim to be an acceptable can- didate (VP-inf). VP b. We rate/consider/*regard/*count Kim an acceptable candidate (NP). (7) c. We rate/consider/*regard/*count Kim quite acceptable (AP). V NP XP d. We rate/consider/*regard/count Kim among the most accepta- ble candidates (PP). e. We rate/*consider/regard/count Kim as an acceptable candidate consider Kim (*as) an acceptable candidate (PP-as). With such a structure, consider can locally constrain the XP ⇒ consider and similar verbs need access to the phrase following Kim Thus, Kim an acceptable candidate doesn’t work as a small clause • But how is the subject of (to be) an acceptable candidate equated with the object of consider ? 9 10 Counter-arguments? Review: ID Schemata (for English): 2 2 ! 3 3 » verb – 4 head ∨ ¬ verb synsem | loc | cat inv − 6 6 7 7 6 5 7 One counter-argument against such structures would be if we had constitu- 6 7 » phrase – subcat �� → 6 7 (Head-Subject) 6 7 dtrs headed-struc 2 3 head-comps-struc ency tests showing that the small clause analysis was best 6 7 6 7 6 dtrs head-dtr phrase 7 6 7 4 5 4 5 ˙ ¸ comp-dtrs sign (8) They consider [Kim quite acceptable], and we consider ? that /? it , 2 ! 3 2 3 » verb – too. 4 head ∨ ¬ verb 6 inv − 7 synsem | loc | cat 6 7 6 7 (9) *It was [Kim quite acceptable] that we considered. 5 ∨ 6 7 (Head-Complement) subcat ˙ synsem ¸ 6 7 6 7 » head-comps-struc – 6 7 dtrs 4 5 head-dtr word In lieu of compelling arguments from constituency tests, either we look at theory-internal arguments or take these local subcategorization facts 2 3 2 3 » verb – 4 head seriously. synsem | loc | cat inv + 6 7 6 7 5 6 7 ∨ subcat �� (Head-Subject-Complement) 6 7 6 7 6 » head-comps-struc – 7 4 dtrs 5 head-dtr word 11 12
Towards an analysis: Unsaturated Complements The empirical challenge of non-finite constructions Where does it say in these ID schemata that every subcategorized item must be realized? In non-finite constructions, the subject of the embedded verb is not expressed as a locally realized dependent. • In English, many verbs and adjectives subcategorize for an unsaturated complement. Problem 1: What is interpreted to be the subject of the non-finite verb? � � � NP � • In other words, a complement can be specified as subcat , rather (10) a. John tried to dance. (subject) b. John promised Peter to dance. (subject) � � than subcat �� c. John persuaded Peter to dance. (object) – The Head-Subject Schema allows for this. – And this will give consider access to the lower subject, as well as its own subject. 13 14 Problem 2: Why do verbs selecting non-finite complements differ w.r.t. Classifying non-finite complements what kind of controllers can occur? (11) a. John tried to dance. (subject) Verbs selecting non-finite complements can be classified b. John appeared to dance. (subject) according to (12) a. John persuaded Peter to dance. (object) • their orientation (subject, direct or indirect object): b. John expects Peter to dance. (object) (13) a. * It tried to rain. (subject) – What is interpreted to be the subject of the non-finite complement? b. It appeared to rain. (subject) (14) a. * John persuaded it to rain. (object) • the nature of the relationship of the embedding verb to the b. John expects it to rain. (object) controller : – Is the controller an argument of the embedding verb? 15 16
Empirical basis of classification The same holds when a dummy subject is permitted : (18) a. There is a dragon in the wood. I. Orientation : Determined by interpretation of embedded infinitive. b. Hobbs is a dragon in the wood. (19) a. There seems to be a dragon in the wood. II. Relationship: embedding verb ↔ controller b. Hobbs seems to be a dragon in the wood. (20) a. * There wants to be a dragon in the wood. Raising verbs only mediate the requirement of the complement: b. Hobbs wants to be a dragon in the wood. 1. If the embedded verb requires a non-referential (= dummy) subject , so does the raising verb: (15) a. It rains. b. * God rains. (16) a. It seems to rain. b. * God seems to rain. (17) a. * It wants to rain. b. * God wants to rain. 17 18 2. If the embedded verb permits a clausal subject , so does the raising 4. In languages where subjectless constructions exist, raising verbs can verb: embed such subjectless complements: (21) a. That Sarah dances fascinates John. (27) a. Dort scheint getanzt zu werden. b. Sarah fascinates John. there seems danced to be (22) a. That Sarah dances seems to fascinate John. ‘People seem to dance over there.’ b. Sarah seems to fascinate John. b. Ihn scheint zu frieren. (23) a. * That Sarah dances wants to fascinate John. him seems to freeze b. Sarah wants to fascinate John. ‘He seems to freeze.’ 3. If the embedded verb has a subject with an idiomatic interpretation , (28) a. * Dort versucht getanzt zu werden. so does it when selected by a raising verb: there tries danced to be b. * Ihn versucht zu frieren. (24) The cat is out of the bag. him tries to freeze (25) The cat seems to be out of the bag. (26) % The cat wants to be out of the bag. 19 20
Recommend
More recommend
