1 Verbal How Questions in MandarinHongyuan Dong Cornell University December 19, 2007 Amsterdam Colloquium Universiteit van Amsterdam, the Netherlands

2 What can we ask? Brutus stabbed Caesar violently?1.argument predicate 4.verb adjunct 1.Who stabbed Caesar violently? 2. What did Brutus do? 3. How did Brutus stab Caesar? 4.Indirectly as: What did John do to Caesar violently? Directly as ?

3 A Mandarin Example Brutus zenme-le Caesar?Brutus how-PERFECTIVE Caesar? What did Brutus do to Caesar? But literally (1) could be translated as: Brutus how-ed Caesar? (This is a direct question, but not an echo-question)

4 Goals and Claims To give a semantic formulation to the verbal “how” questions. I claim that the verbal “how” ranges over properties of events. To give a compositional semantics of such questions. I propose a restricted variable approach and a corresponding abstraction rule. To account for three restrictions on the use of such questions. I argue that the verbal “how” is uniformly used as a transitive verb, and that the patient role should be further distinguished by the [affective] feature to trigger the malefactivity presupposition. To evaluate two approaches to indefinites: choice functions (Reinhart 1998) vs. Structured Variable (Abusch 1994). I argue that a semantic account of the verbal “how” questions can be achieved in the Structured Variable approach , but not in the Choice Function approach. To discuss the connection between the meanings of the verbal “how” and the adverbial “how”. I point out that they are both related to properties of events.

5 Basic Data 1. Zenme-le? 3. Yuehan zenme-le Bi’er? John how-ed BillWhat did John do to Bill?4. 4.*Yuehan zenme-le Bi’er yi-ben-shu John how-ed Bill one-book John what Bill a book? 1. Zenme-le? How-ed What happened? 2. Yuehan zenme-le? John how-ed What happened to John?

6 Three Restrictions John how-ed? What happened TO John?Non-Agentivity: John how-ed? What happened TO John? ≠ What did John do? Non-Transitiviy: * John how-ed Bill a book?

7 Three Restrictions John kissed Mary, and she was happy.3. Malefactivity: John kissed Mary, and she was happy. * John how-ed Mary? John kissed Bill, and he was annoyed. √ John how-ed Bill?

8 THEORETICAL BACKGROUNDPartition Semantics Intuitively, a person who asks a question wants to be relieved from a state of ignorance with respect to a certain piece of fact of the world, and they want to differentiate between all the possibilities and try to figure out which one is real.

9 I only drew 7 for simplicityTHEORETICAL BACKGROUND w2, w18, … w27, w7,… w56,… w92,… w101, … w73,… w20, w42, w58, … 16 cells altogether, I only drew 7 for simplicity Who went to the party? Set of individuals: {Adam, Bill, Chris, Dan} partition of the set of possible worlds 16 cells, each of which corresponds to one possibility λw λw’[λx .person(w0)(x) ∧ party’(w)(x) = party’(w’)(x)]

10 Proposition-set SemanticsTHEORETICAL BACKGROUND Proposition-set Semantics The denotation of a question is the set of propositions that are possible answers.

11 { that Adam went to the party; that Bill went to the party; THEORETICAL BACKGROUND Who went to the party? { that Adam went to the party; that Bill went to the party; that Chris went to the party; that Dan went to the party. } c. λp∃x [person’(w)(x) ∧ p = λw’. party’(w’)(x)]

12 An Event Semantics for the verbal HowWhat are good answers? a. Yuehan da-le Bi’er. John hit-ed Bill John hit Bill. b. Yuehan da-le Bi’er, ye ma-le Bi’er. John hit-ed Bill, also scold-ed Bi’er John hit Mary and also scolded Mary. How do we give the semantic representation of such sentences? How do we characterize the verbs in these answers

13 Event Semantics Davidson (1967) & Parsons (1990)Brutus stabbed Caesar violently. ∃e [ Stabbing(e) ∧Agent(e, Brutus) ∧Patient(e, Caesar) ∧Violent(e)] The main verb is treated as a property of events.

14 Therefore: a. {that John hit Bill, that John scolded Bill, ……} b. {p|∃P. [p=^∃e. [P(e) ∧Agent(e, John) ∧Patient (e, Bill)]]}

15 Compositional SemanticsThe goal is to derive the semantics in a compositional way Yuehan zenme-le Bi’er {p|∃P. [p=^∃e. [P(e) ∧Agent(e, John) ∧Patient (e, Bill)]]} Two Problems to Solve: Derive the question semantics compositionally Derive the event semantics compositionally

16 Derivation of “who” questions (Lahiri 2002)CP NP C’ C IP who ? t1 went-to-the-party IP: party’(w)(x1) C1: λpλq [ q = p] C’: λx1λq [ q = λw. party’(w)(x1)] NP : λQλp∃x [ person’(w)(x) ∧ Q(x)(p)] CP : λp∃x[ person’(w)(x) ∧q = λw’. party’(w’)(x)]

17 So does the verbal how move?Wh-in-situ and movement in Mandarin: Wh-arguments: do not move Wh-adjuncts: move Evidence: Island Escaping Ability Wh-arguments: yes Wh-adjuncts: no Thus: if the verbal how can escape islands, it does not move.

18 a. Yuehan xihuan shei xie de shu?John like who wrote DE book [whoi [John likes the book whoi wrote]] b.*Yuehan xihuan ni zenme xie de shu John likes you how wrote DE book *[how [John likes the book that you wrote how]] c. Yuehan xihuan Mali zenme-le de ren? John like Mary how-ASP DE[1] person. [how [John likes the person that Mary how-ed]] [1] DE is a structural morpheme in relative clause constructions in Mandarin Chinese.

19 Thus the verbal how does not move, and shouldbe bound by a default Q morpheme. [Q [ … how… ] ] Lahiri’s semantics only applies to: [NPi ? [ …NPi…]] We need a semantics for the Q morpheme here: Berman’s (1994) rule for the Q morpheme 〚Qφ〛M,g = {p: ∃(x1…xn) [p=〚Qφ〛M,g’ ]}, where g’≈φ g. domain restriction in-situ

20 Who will be offended if we invite which philosopher? A problem for Berman’s rule: Who will be offended if we invite which philosopher? for which , if we invite y and y is a philosopher, then x will be offended. Reinhart (1998) domain restriction cannot be in-situ.

21 My temporary solution Restricted variables xDVariables that carry their domain restriction with them for abstraction by relevant rules. Abstraction rule If α is a branching node, {β, γ} is the set of α’s daughters, β is an unselective binder Op, and γ contains a restricted variable xD, then 〚α〛= Op x such that x∊D.〚γ(…x…)〛 Q Morpheme (Lahiri + Berman + abstraction) 〚Q〛=λq.λp.∃x∊D.p=q, where q contains a restricted variable xD

22 Now we move on to the second problem to see how we can derive the event semantics compositionally.use Kratzer’s (1996) event-identification rule

23 VoiceP and Derivation (Kratzer 1996: 121)a. Mittie feed the dog (untensed sentence) b. structure of (36a) VoiceP DP Voice’ Mittie Voice VP Agent DP V the dog feed

24 c. VoiceP: semantic interpretation[1] feed* = λxe λes [feed(x)(e)] [2] the dog* = the dog [3] (the dog feed) * = λes [feed(the dog)(e)] (From (1), (2) by Functional Application) [4] Agent* = λxe λes[Agent(x)(e)] [5] (Agent (the dog feed))* = λxe λes[Agent(x)(e) & feed(the dog)(e)] (From (3), (4) by Event Identification) [6] Mittie* = Mittie [7] ((Agent (the dog feed)) Mittie)* =λes[Agent(Mittie)(e) & feed(the dog)(e)] (From (5), (6) by Functional Application)

25 Event Identification (Kratzer 1996: 122)f g  h > > λxe λes[f(x)(e) & g(e)]

26 The result of the untensed derivation in the previous two slides is a property of events, which can be (existentially) closed by aspect or tense, e.g. the Chinese –le. Lin (2004) provides an interpretation for the –le, which I will use here. Now finally we have solved both problems and we are in a position to do the compositional

27 Compositional DerivationSample derivation of “Zhangsan zenme-le Lisi?” CP Q TP DP T’ T AspP F vP -le VoiceP Zhangsan Voice’ Agent VP Lisi zenme f1D>

28 Compositional DerivationDerivation steps[3] [1]〚zenme〛=λx. λe. λw. f1D> (x)(e)(w). [2]〚VP〛= λe. λw. f1D> (Lisi)(e)(w) [3]〚Agent〛= λx. λe. λw. Agent(x)(e)(w) [4]〚Voice’〛= λx. λe. λw. [f1D> (Lisi)(e)(w) ∧Agent(x)(e)(w)] [5]〚VoiceP〛=λe.λw.[f1D>(Lisi)(e)(w)∧Agent(Zhangsan)(e)(w)] [6]〚-le〛= λt. λe. λw. [F(t)(e)(w)] [7]〚vP〛= λt. λe. λw. [f1D> (Lisi)(e)(w) ∧Agent(Zhangsan)(e)(w) ∧F(t)(e)(w)] (continued on next slide) [3] All of the type labels are the same as before. The variable w ranges over possible worlds. I also use the letter w as the type label for worlds, since the usual label s has already been used as that for events.

29 Compositional Derivation[8]〚F〛= λt. λw. ∃e [F(t)(e)(w)] [9]〚AspP〛= λt. λw. ∃e [f1D> (Lisi)(e)(w) ∧Agent(Zhangsan)(e)(w) ∧F(t)(e)(w)] [10]〚T〛= t0 (Speech Time) [11]〚TP〛= λw. ∃e [f1D> (Lisi)(e)(w) ∧Agent(Zhangsan)(e)(w) ∧F(t0)(e)(w)] [12]〚Q1〛=λq. λp. ∃x such that x∊D. p=q [13]〚CP〛= λp. [∃f such that f ∊ D>. p= λw. ∃e [f(Lisi)(e)(w)∧Agent(Zhangsan)(e)(w) ∧F(t0)(e)(w)]]

30 Rule of Event Closure (Lin 2004:629)[4] Notes: Rule of Event Closure (Lin 2004:629)[4] + > Condition: F must match the aspect feature F introduces the time dimension as in F(t0)(e)(w), which says the event e precedes the speech time in world w. in both Kratzer’s (1996) and Lin’s (2004) derivations, they only deal with the extension of a sentence, while in my derivation in order to derive the semantics of questions as sets of propositions, I have added the world variable all along. [4] The label i stands for time instances/time intervals.

31 Explanation of the three restrictionsClaim: the verbal “how” is uniformly used as a transitive verb, i.e. fD> 〚zenme〛=λx. λe. λw. fD> (x)(e)(w).

32 Non-Agentivity Zhangsan zenme-le? CP Q1 TP DP T’ T AspP F vP -le VPf1D>

33 [2]〚VP〛= λe. λw. f1D> (Zhangsan)(e)(w) [1]〚zenme〛=λx. λe. λw. f1D> (x)(e)(w). [2]〚VP〛= λe. λw. f1D> (Zhangsan)(e)(w) [3]〚-le〛= λt. λe. λw. [F(t)(e)(w)] [4]〚vP〛= λt. λe. λw. [f1D> (Zhangsan)(e)(w) ∧F(t)(e)(w)] [5]〚F〛= λt. λw. ∃e [F(t)(e)(w)] [6]〚AspP〛= λt. λw. ∃e [f1D> (Zhangsan)(e)(w) ∧F(t)(e)(w)] [7]〚T〛= t0 (Speech Time) [8]〚TP〛= λw. ∃e [f1D> (Zhangsan)(e)(w) ∧ F(t0)(e)(w)] [9]〚Q1〛=λq. λp. ∃x such that x∊D. p=q [10]〚CP〛= λp. [∃f such that f∊D>. p= λw. ∃e [ f(Zhangsan)(e)(w) ∧F(t0)(e)(w)]]

34 Non-Transitivity 〚zenme〛=λx. λe. λw. f D> (x)(e)(w). There is no position for a third argument. For the zero-place use of “how”, the subject is dropped.

35 f1D>a. [ e zenme-le] b. Tree structure CP Q TP e T’ T AspP F vP -le VP e zenme f1D>

36 Malefactivity It is a non-cancelable presupposition of the construction Where is it triggered? Sever the internal argument from the verb, as Lin’s (2004) paper does, and further distinguish different patient roles with the [affective] feature.

37 VoiceP Zhangsan voice’ Agent vp Lisi v’ Patient V [+aff] zenme

38 Choice Functions vs. Structured VariableThe indefinite/wh-in-situ dilemma: a. Who will be offended if we invite which philosopher? b. for which , if we invite y and y is a philosopher, then x will be offended. c. for which , y is a philosopher, and if we invite y, x will be offended. Task: a) bind the variable non-locally; b)extract the domain restriction

39 Choice Function (Reinhart 1998)For which , (CH(f) ∧(we invite f(philosopher) x will be offended)). *Who arrived how?

40 Reinhart’s (1998) explanation:One thing that would be agreed upon in all frameworks is that wh adverbials are different from wh-NPs. First, because they do not have an N-set, hence no N-role or variable; and second, because they denote functions ranging over higher-order entities (Szabolcsi and Zwarts 1993). This entails that they cannot be interpreted via choice functions selecting an individual from a set (since there is neither a variable that can be bound by forming a set nor a set of individuals that the choice function could apply to). ——Reinhart (1998: 45)

41 1. Choice functions only apply to wh-in-situ2. Choice functions do not apply to higher-order entities. 3. English “how” has to move and should be interpreted by other means. Could there be a “how-in-situ” language?

42 Mandarin Chinese Zenme “how”Adverbial “how” in Mandarin has to move, although wh-arguments do not move. This is a nice result of Reinhart’s theory 2. But the verbal “how” does not move. Are they uninterpretable according to Reinhart’s choice function theory?

43 Interpretation dilemma (for wh-in-situ in Chinese):Wh-argument: via choice functions; Wh-adjunct: probably via Lahiri’s semantics for moved wh-elements; Verbal-how: no interpretation?

44 Structured Variable (Abusch 1994)General schema “φ: U”, where U is a set of indices of unquantified in-situ indefinites. a. arrive(x1):{} (indefinites) b. arrive(x1):{ } (definites)

45 a A man arrived and a woman left b. representation < x1,man(x1)> arrive(x1) ∧ leave(x2) : < x2,woman(x2)> arrive(x1) : {< x1,man(x1)>} leave(x2): {< x2,woman(x2)>}

46 Existential Closure Rule: (Abusch 1994)Where ,…,< xkn,φn > ∈2(X’), the interpretation is: ∃xk1…∃xkn [φ1∧…∧φn∧1(X’)] : [2(X’) – {,…,< xkn,φn > }] Note: 1(φ: U)= φ , 2(φ: U)=U

47 The Chinese verbal “how” can be interpreted by the Structured Variable approach. (My abstraction rule can be considered an extension of this approach to questions) But this approach does not deal with the how-in-situ in English

48 A final comparison If we take these assumptions:The problem with in-situ “how” in English is semantic; The semantic interpretation mechanisms are meant to be applicable to different langauges. Then we have the following pros and cons: The verbal “how” is a problem for the Choice Function approach, but not for the Structured Variable. The in-situ “how” in English is still to be explained, if we adopt the structured variable approach for all wh-in-situ.

49 A different explanation?If the problem of the English in-situ “how” is syntactic (Diesing 1993): There is no reason why choice functions cannot be applied to sets of higher-order entities. The two approaches have equal interpretive power.

50 A Broader Picture Adverbial “how”: How did John dance?John danced beautifully. ∃e [ Dancing(e) ∧Agent(e, John) ∧Beautiful(e)]

51 “How” as properties of eventsBoth the adverbial how and the verbal how range over properties of events. Adverbial how: manner properties? Verbal how: type of events?

52 Structure of Manners Szabolcsi, A and Zwarts, F. 1993.Manners range over domains of free join semi-lattice structure. “How” as properties of events: {e1, e2, e3, e4, …} Is there a way of structuring the set of events, or the set of sets of events?

53 Remaining Issues Q. Yuehan zenme-le? John how-ed?What happened to John? A: He tripped over a rock. Indirect answer?  “Zenme” is not limited to a transitive verb?

54 “What” Questions What did John do to Bill? Yuehan zenme-le Bi’er.Similar to the verbal “how” questions?

55 Any other languages? Is this type of verbal how questions available in other typologically related languages? Wh-in-situ (Is this type of questions imaginable in English?) Separate aspect morphemes (e.g. –le in Mandarin)

56 Conclusions The verbal how ranges over properties of events. It is uniformly used as a transitive verb. The requirement for the “how” to move in some environment but not in others poses a problem for the choice function approach to wh-in-situ, but this fact supports the structured variable approach. Both the verbal “how” and the adverbial “how” range over properties of events

57 Thank You!

58 Essential References Abusch, Dorit The scope of indefinites. Natural Language Semantics 2: Kluwer. Baker, C.L “Notes on the description of English questions: the role of an abstract question morpheme,” Foundations of Language 6: Berman, S On the Semantics of Wh-Clauses. Garland, New York and London. Davidson, Donald The Logical Form of Action Sentences. In N. Rescher (ed.), The Logic of Decision and Action, University of Pittsburgh Press. Diesing, Molly “Multiple Multiple Questions”, in Andrew Carnie, Heidi Harley and Mary Willie (eds.) Formal Approaches to Function in Grammar, John Benjamins, Hamblin, C. L Questions in Montague English. Foundations of Language 10:41-53. Heim, Irene The Semantics of Definite and Indefinite Noun Phrases. New York: Garland. Higginbotham, James, and Fabio Pianesi, Achille C. Varzi. (eds) Speaking of Events. Oxford University Press. Huang, James Logical Relations in Chinese and the Theory of Grammar. PhD dissertation, MIT, Cambridge, MA.

59 Karttunen, Lauri. 1977. Syntax and Semantics of QuestionsKarttunen, Lauri Syntax and Semantics of Questions. Linguistics and Philosophy 1:3-44. Reprinted in Portner, Paul and Partee, Barbara H. (eds) (2002). Kratzer, Angelika Severing the external argument from its verb, in Rooryck, Johan and Zaring, Laurie (eds) Phrase Structure and the Lexicon. Kluwer. Lahiri, Utpal Questions and Answers in Embedded Contexts. Oxford University Press. Lin, Tzong-Hong Jonah Aspect, Distributivity, and Wh/QP Interaction in Chinese. Language and Linguistics 5.3: Taipei, China Lewis, David Adverbs of Quantification. In E. Keenan (ed.) Formal Semantics of Natural Lanuage, 3-15, Cambridge University Press, Cambridge, UK. McCloskey, James Quantifier Float and Wh-Movement in an Irish English. Linguistic Inquiry 31: 57-85 Parsons, T Events in the semantics of English: A study in subatomic semantics. Cambridge, MA: MIT Press. Reinhart, Tanya Wh-in-situ in the Framework of the MiniBi’erst Program. Natural Language Semantics 6:

60 Reis, Marga. 1992. The category of invatiant alles in wh-clauses: On syntactic quantifiersvs. quantifying particles in German. In Who climbs the grammar tree?,ed. Rosemarie Trace Tuebingen: Max Niemeyer Verlag. Shimoyama, J Indeterminate Phrase Quantification in Japanese. NaturalLanguage Semantics, 14: Szabolcsi, A and Zwarts, F Weak Islands and an Algebraic Semantics for Scope Taking. Natural Language Semanics. Tsai, Wei-Tien Dylan On Economizing the Theory of A-BarDependencies. Garland Publishing, Inc. Xu, Liejiong Free empty category. Linguistic Inquiry 17: