Linguistic orphans: composition or construction?∗Line Mikkelsen (UC Berkeley), CBS, October 24, 20111 OrphansOvert VP anaphors (do so/it/the same) allow preposition-marked orphans with antecedent-internal correlates:(1) You have jilted two previous fiances and I expect you would do the same to me.anaphor do the sameorphan meantecedent jilted two previous fiancescorrelate two previous fiancesChallenges for syntax-semantics mapping:• Correlate is inside antecedent VP ⇒orphan and antecedent must interact to produce interpretation: and I expext you would jilt me• Orphan and correlate are both interpreted as Patient arguments, but differ syntactically: PP vs. NPToday’s question Is ‘do the same to X’ a construction—i.e. a fixed pairing of form and meaning—or is itcomposed by general syntactic and semantic rules?Larger questions1. To what extent is syntax autonomous and general?2. To what extent is semantics compositional?3. What units are manipulated by syntax?• Mainstream generative tradition (Chomsky 1957, Montague 1974 and onwards)1. Syntax is fully autonomous and general.2. Semantics is fully compositional (rule isomorphism, only sisters compose).3. Syntax manipulates words (and bound morphemes).• Construction Grammar (Fillmore 1985), Sign-Based Construction Grammar (Boas and Sag to appear),Simpler Syntax (Culicover and Jackendoff 2005).1. Syntax is partly autonomous and somewhat particular.2. Semantics is partly compositional.3. Syntax manipulates constructions, which can be of any size (morphemes, words, phrases, supra-sentential chunks).∗The first stage of this work was made possible by a shared grant from the Danish Agency for Science, Technology, and Innovation.It has benefitted from exchanges with Russell Lee-Goldman, Dan Hardt, Caroline Heycock, Ray Jackendoff, Andy Kehler, JennyLederer, Jason Merchant, Ivan Sag, Bjarne Ørsnes and audiences at UC Berkeley, Copenhagen Business School, Stanford, andUniversity of Arizona, Tucson. I am also grateful to the three Berkeley undergraduate students, Erik Maier, Nicholas Strzelczyk andMelissa Valdes, who have been populating the orphans database with example from corpora.12 Constructional analysis of orphans• How do we recognize a construction when we see one?[C]onstructions [are] often endowed with properties which are not independently determined byfacts about their constituency or their derivation. (Fillmore 1985:73)• Some proposed constructions (proposed by/deconstructed by):1. the heck intrusion (Fillmore 1985/den Dikken and Giannakidou 2002)(2) What the heck did you fix it with?2. let alone (Fillmore et al. 1988/Toosarvandani 2010)(3) I doubt you could get Fred to eat squid, let alone Louise.3. What’s X doing Y? (Kay and Fillmore 1999/ ?)(4) What is this scratch doing on the table?4. The X-er, the Y-er. (Culicover and Jackendoff 1999/den Dikken 2005)(5) The more you eat, the less you want.5. Bare Argument Ellipsis (Culicover and Jackendoff 2005/Merchant 2004)(6) Q: Has Harriet been drinking again?A: Yeah, scotch.• Possible constructional properties of ‘do X to Y’1. Verb must be main verb do; auxiliaries don’t work:(7) You have jilted two previous fiances and . . .a. *I suspect he has the same to his.b. *I suspect you will the same to me.2. Orphan NP has argument semantics—patient, theme, or benefactive—, but not argument syntax: PPinstead of NP:(8) You have jilted two previous fiances and I expect you would do the same *(to) me. [patient](9) North Korea has sold medium-range SCUD missiles to Iran and Syria and it could do thesame *(with) the Rodong. [theme](10) If you hope people will re-tweet what you write, you have to do the same *(for) other people.[benefactive]3. Orphan interpretation is guided by correlate, but correlate need not be an argument (or even depen-dent!) of the antecedent verb:(11) And he banged the receiver in her ear before she could do the same to him.(12) If you hope people will re-tweet what you write, you have to do the same for other people.2Construction rule (based on Culicover & Jackendoff’s (2005:298) rule for orphans with do so):(13) Syntax: [vp[vdo][npthe same] hPPorphii] Conceptual Structure:[ActionF(. . . ); . . . hYii. . . ]where:a. hXi marks X as optionalb. F is an open function whose content is filled in by the antecedent (indirect licensing)c. PP is the orphan and it is connected to a correlate via the optional semantic constituent Y in thedomain of FThe rule in (13) specifies1. the verb as do, ruling out auxiliaries2. the orphan as a PP, ruling out NPs3. the correlate semantically; no syntactic restriction on correlateThematic contribution of orphan P not expressed, but could be done by linking identity of P to interpretation ofY:• . . . PforPi. . . hYbenefactiveii• . . . PtoPi. . . hYpatientii• . . . PwithPi. . . hYthemeii3 A compositional analysisHardt et al. (2011) propose a compositional analysisSyntax The orphan is an adjunct to VP(14) VPXXXXXVPQQVdoNPbb""the samePPHHHPto/with/forDPBBXSemantics The presence of the orphan adjunct forces lambda abstraction over the correlate in the antecedent,creating a “slot” for the orphan in the meaning reconstructed for the anaphor.Step 1. abstraction [jilt two previous fiances] ⇒ [jilt x]Step 2. reconstruction do the same to me ⇒ [jilt x] (me)Step 3. function application [jilt x] (me) ⇒ [jilt me]This analysis is compositional because:• adjunction is a general syntactic mechanism• lambda abstraction is a general semantic mechanism (used e.g. in interpretation of relative clauses andsloppy identity)34 Advantages of a compositional analysisThe following observations are explained by the compositional analysis, but must be stipulated in the construc-tional analysis:1. P-marking It is characteristic of adjuncts to VP that they are PPs and not NPs.1If orphans are adjunctsto VP, we expect them to be PPs and not NPs:(15) You have jilted two previous fiances and I expect you would . . .a. do the same to me.b. *do the same me.2. Optional ity It is characteristic of adjuncts that they are optional. If orphans are adjuncts, we expect themto be optional and they are:(16) He folded up his jacket and sat on it. I did the same (with mine).3. Iteration Another characteristic of adjuncts is that they may iterate. If orphans are adjuncts, we thusexpect multiple orphans to be possible and they are:(17) I convened convened a seminar [corr. 1at the beginning of June] to
or
We will never post anything without your permission.
Don't have an account? Sign up