The Movement Of Question Particles

The movement of question particles 3. Evidence, part 2: Island effects in Japanese—or a lack thereof Islands block movement, yet (6) Japanese allows wh-words in islands… 1. (1) 2. The proposal dare-ga tka hon-o kaimasita book-ACC bought.POLITE who-NOM ‘Who bought a book?’ ka ? Q The question particle undergoes syntactic movement from a clause-internal position (by the wh-word) to the clause periphery (in the complementizer system). Evidence, part one: Intervention effects in Japanese Assume: Movement obeys Attract Closest. Scheme: Put an eligible alternative to Q along the path of movement. 3 X … 3 " 1 Y … 1 3 … z--=-- tX This should interfere with the movement. Some intervenors for attraction of ka ‘Q’: • disjunction particle ka • indefinite-forming particle ka (2) (3) Attract Closest: Movement is motivated by the need to check a feature. Only the closest element with the relevant feature is eligible for movement. (Chomsky 1995) So if ka is moving, it must be moving from outside… (7) Hiro-ga [ Sue-ni nani-o ageta hito-ni ] aimasita ka? H-NOM S-DAT what-ACC gave man-DAT met.POL Q ‘*What did Hiro meet [the man that gave t to Sue]?’ [island … nani … ] … tka … ka ? Whoa—does this mean islands are useless for diagnosing movement? Well, no—but we have to be more clever. Suppose: Emphatic ittai locates the launching site of ka. This isn’t crazy. ka can sometimes be dropped, but never with ittai. (8) John-ga ittai tka nani-o kaimasita ka? John-NOM ittai what-ACC bought.POL Q ‘What in the world did John buy?’ (9) ittai inside an island ⇒ ka moved out of the island. (11) ill-formed as expected. (10) ?? Hiro-ga ittai nani-o tabeta? H-NOM ittai what-ACC ate (‘What in the world did Hiro eat?’) Hiro-ga nani-o tabeta? H-NOM what-ACC ate ‘What did Hiro eat?’ ittai outside an island, ka could move from outside. (12) is fine. Y INTERVENES for movement of X if Y is equally eligible to move, since Y is closer to the target of movement than X was. John-ka Bill-ga hon-o kaimasita. John-or Bill-NOM book-ACC bought.POLITE ‘John or Bill bought a book.’ (11) * Hiro-ga [ Sue-ni ittai nani-o ageta hito-ni ] aimasita ka? H-NOM S-DAT ittai what-ACC gave man-DAT met.POL Q (‘What in the world did Hiro meet the man that gave t to Sue?’) (12) ageta hito-ni ] aimasita ka? Hiro-ga ittai [ Sue-ni nani-o H-NOM ittai S-DAT what-ACC gave man-DAT met.POL Q ‘What in the world did Hiro meet the man that gave t to Sue?’ 4. Combining islands and intervention effects dareka-ga hon-o kaimasita. someone-NOM book-ACC bought.POLITE ‘Someone bought a book.’ And, indeed: Another -ka cannot be on the path of Q-movement (noticed by Hoji 1985) (4) ?* John-ka Bill-ga nani-o kaimasita ka? John-or Bill-NOM what-ACC bought.POLITE Q (‘What did John or Bill buy?’) (5) nani-oi John-ka Bill-ga ti kaimasita ka? what-ACC John-or Bill-NOM bought.POLITE Q ‘What did John or Bill buy?’ Q attracted from below the subject—thus, ka ‘or’ intervenes. Idea: • Independent -ka blocks Q-movement (“intervention effect”) • Q-movement takes place from outside of islands. • Therefore, Q-movement should be insensitive to intervenors inside island Q scrambles (with nani) above intervenor. From there, attraction is unimpeded. 1 2 Embed “intervention effect” in an island and it becomes well-formed, cf. (4) on previous page. The idea: So: (13) Mary-wa [ John-ka Bill-ga nani-o katta ato de ] Mary-TOP John-or Bill-NOM what-ACC bought after (14) Mary-wa [ nani-oi John-ka Bill-ga ti katta ato de ] Mary-TOP what-ACC John-or Bill-NOM bought after ‘Mary left after John or Bill bought what?’ 5. Ok, maybe you’re right about Japanese, but so what? Well, this isn’t just about Japanese. Chitra mokak d\ gatte? Chitra what Q bought-E ‘What did Chitra buy?’ Chitra mokak d\ gatta. Chitra what Q bought ‘Chitra bought something.’ 6. In both languages, there is a movement relation involving Q. Evidence for Q-movement in Sinhala First of all, sometimes d\ appears overtly at the clause periphery in Sinhala— but at the expense of -e morphology on the verb. Premodern Japanese and (Shuri) Okinawan look a lot like Sinhala in these respects too. (17) Sinhala and Japanese are showing us two sides of the same movement. Q starts where we see it in Sinhala and winds up where we see it in Japanese. Question word remains in situ Question word followed by Q Verb takes special “-e” form Sinhala d\ (16) corresponds to Japanese ka (17): (16) But d\ sits inside the clause, while Japanese ka sits at the periphery. dekakemasita ka? left.POLITE Q • “Q-movement” happens in other languages. • There may be a semantic motivation for it. Sinhala: Indo-European (Sri Lanka), SOV language with scrambling, wh-in-situ . (15) dekakemasita ka? left.POLITE Q We have identified d\ as an analog in Sinhala to Japanese ka. Reason one: Wh+Q = Indefinite (22) Ranjit [ kau d\ aawe kiy\la ] dann\wa. Ranjit who Q came-E that know ‘Ranjit knows who came.’ d\ … aawe (23) Ranjit [ kauru aawa d\ kiy\la ] dann\wa. Ranjit who came Q that know ‘Ranjit knows who came.’ … aawa d\ dare-ka-ga hon-o kaimasita. who-Q-NOM book-ACC bought.POL ‘Someone bought a book.’ Also: notice that the -e morphology only appears when d\ is not after the verb. That -e morphology determines the scope of the question word Sinhala d\ (18) corresponds to Japanese ka (19): Reason two: declarative+Q = y/n question (18) Taroo-ga hon-o kaimasita ka? Taroo-NOM book-ACC bought.POL Q ‘Did Taro buy a book?’ Chitra ee pot\ gatta d\? Chitra that book bought Q ‘Did Chitra buy that book?’ (19) Sinhala d\ (20) corresponds to Japanese ka (21): Reason three: Q used for disjunction (21) (20) John-ka Bill-(ka-)ga hon-o katta. mahatteat.\ tee d\ koopi d\ oon\? John-Q Bill-(Q-)NOM book-ACC bought gentleman-DAT tea Q coffee Q necessary ‘Do you (sir) want tea or coffee?’ ‘John or Bill bought books.’ (24) Ranjit [ kau d\ aawe kiy\la ] dann\wa. Ranjit who Q came-E that know ‘Ranjit knows who came.’ -e marks embedded verb ⇒ embedded question (25) Ranjit [ kau d\ aawa kiy\la ] danne? Ranjit who Q came that know-E ‘Who does Ranjit know came?’ -e marks matrix verb ⇒matrix question What could look more like feature-driven movement? Idea: -e is a morphological reflex of the feature which will drive movement (attraction) of Q. Where Q goes is tied to its semantic scope. 3 4 So But Conclusion: Q can’t start by the higher wh-word, it must start by the lower one. d\ in Sinhala and ka in Japanese are the same thing, following the same route. we can see in Sinhala where Q moves from, something we can’t see in Japanese. Caveat: The data is fairly shaky here. It is also possible to have d\ by both wh-words. Sumangala (1992) suggests that (30) may in fact be derived via ellipsis from (31), citing as evidence that both cannot be answered with a list of pairs, whereas (29a) can be. Perhaps that’s right, but it requires more investigation. Well when a wh-word is inside an island, does Q really move from outside? We deduced that it must from looking at Japanese—but in Sinhala we can see it directly: Sinhala: Question words allowed in islands—but only if d\ is immediately outside the island. (26) * [ Chitra mon\wa d\ kan\ kot.\ ] Ranjit pudum\ unee ? Chitra what Q ate when Ranjit surprise became-E (‘Ranjit was surprised when Chitra ate what?’) (27) [ Chitra mon\wa kan\ kot.\ ] d\ 7. Multiple questions Ranjit pudum\ unee ? Chitra what ate when Q Ranjit surprise became-E ‘Ranjit was surprised when Chitra ate what?’ Look… Sinhala provides overt evidence for the syntactic structure we could only infer in Japanese. 7. (30) kau d\ mon\wa d\ kieuwe? who Q what Q read-E ‘Who read what?’ (requires stress on both kaud\ and mon\wad\) (31) kau d\ kiewe mon\wa d\ kiewe? who Q read-E what Q read-E ‘Who read, what did s/he read?’ Q and the pair-list question Suppose we force the situation in Japanese by putting two wh-words inside an island. We know Q can’t get out of an island—it can’t start by either wh-word in this case. Rather, it will have to start outside of the island. If Q “starts by the wh-word,” what happens in multiple questions? Specifically, there’s only one ka in (28). Where does it move from? By which wh-word? (32) (28) dare-ga nani-o kaimasita ka? who-NOM what-ACC bought.POLITE Q ‘Who bought what?’ Japanese (32) is grammatical— but can only be answered with a single pair of person and purchased item. Again, we can turn to Sinhala to help answer this question— (29) a. [kauru mokak d\ kieuwe kiy\la] dann\wa d\? who what Q read-E that know Q ‘Do (you) know who read what?’ b. * [kau d\ mokak kieuwe kiy\la] dann\wa d\? who Q what read-E that know Q (‘Do (you) know who read what?’) Taroo-ga [ dare-ga nani-o katta toki-ni ] okotta no? got.angry Q Taro-NOM who-NOM what-ACC bought when ‘Taroo got angry when who bought what?’ (*PL, SP) Sinhala It appears that having Q start out by the lower wh-word is crucial to getting the PL reading. How is it that Q has to start with the lower wh-word? Why must it be attached there? Speculative answer: It’s an economy condition, which works like this: If a wh-word is merged into the representation (the structure being built from the bottom up) and a Q is available in the “numeration”, merge it immediately. The intuition: Introducing a wh-word causes a great deal of “work” until Q is merged. 5 6 That’s pretty vague—what is Q anyway? And what’s so hard about a wh-word? 8. So… Recall that Q appears (16–21) in several contexts. It is not directly linked to interrogativity: • questions in Japanese need not have ka—recall (9). • ka can appear in non-questions—e.g. in the declarative (17) (with dareka ‘someone’). Q-movement… A rapid sketch of the semantics of Q and wh-words developed in Hagstrom (1998). So, it may in fact be a general fact about language(s)—even where you can’t see it happening. (See Bošković 1998 for an attempt to extend this to English). Following Hamblin (1973), • interpret a wh-word as a set of individuals. • interpret a question as a set of propositions. • allow function application to be “flexible” • seems to happen in several—unrelated—languages. • may be a crucial part of the semantics of questions and indefinites. Roughly: For some set A containing elements of type α and for some function f from type α, f(A) is interpreted as being the set λx∃a∈A.x=f(a). Since dare ‘who’ and ka ‘Q’ are both in wh-questions and in indefinites (like dareka ‘someone’), and [[ dare ]] is a set of individuals, what must ka mean? Proposal: Q introduces existential quantification over choice functions. So: and [[ who left? ]] comes out to be the set λp.∃f.p=f( [[ who ]] ) left, [[ someone left ]] comes out to be the proposition ∃f.f( [[ who ]] ) left. Syntactic movement of Q creates a chain. The syntactic location of the top of the chain (where Q moves to) determines the scope of the existential quantifier ( ∃f ), and the syntactic location of the bottom of the chan (where Q starts from) determines where the choice function variable ( f ) is (e.g., by who). Now: The “work” involved in a wh-word comes from the “flexible” functional application. If a function which takes something of type α instead gets a set of things of type α, the evaluation is carried out on each member of the set individually. So, you have a set of resulting outputs of the function instead of a single output. Q introduces a choice function, which takes a set and chooses one member. After Q is introduced, the type is resolved and flexible functional application is no longer needed. Consider combining what with transitive bought. Without Q, the result of combining them would be a set of predicates λP∃x[P=bought(x) ∧ x∈ [[ what ]] ], whereas if Q introduces a choice function variable as a sister of what, you would have a single predicate (with an unbound variable which will be bound by Q in its landing site) bought(f( [[ what ]] )). If flexible functional application is “costly”, there is motivation to introduce Q as soon after a wh-word as possible. 7 8