Notes on Conditions

Conditions... "...On Transformations" 2. SSC/TSC 1. What are conditions on transformations? Passive should apply blindly to: (4) a. I believe the dog to be hungry. b. I believe the dog is hungry. ƒ A/A (p.85) as a condition on application of transformations. More precisely, a condition on the choice of factorization in case of ambiguity. Passive SD: SC: (1) X NP V NP Y 1 2 3 4 5 1 4 be - en 3 by+2 5 But it doesn't: (5) a. The dog is believed to be hungry. b. *The dog is believed is hungry. (6) Tensed S Condition: version 1 No rule can involve X, Y in the structure . . .X . . .[α . . . Y . . . ] where α is a tensed sentence. [20] A factorization blocked by A/A John's winning — the race — surprised — me — ø ==> 1 2 3 4 5 *John's winning I was surprised by the race. ƒ The TSC derives effects of the "Insertion Prohibition" (= phase impenetrability for insertion of "morphological material") for tensed sentences. ƒ An example of a condition on the functioning of a specific rule. ƒ ƒ One can also imagine conditions on the form of rules, i.e. conditions on the grammars themselves. Assumption: Each other derived by each lowering The candidates each hated the others. ==> The candidates hated each other. (2) Comp-substitution universal. No rule may fill COMP by rightward movement. (3) Background on Comp: C empty in the base. Filled by transformation: o with wh if possible, (7) a. The candidates expected each other to win. b. *The candidates expected that each other would win. (8) Overt subjects blocking each-movement: towards SSC a. The candidates each expected [PRO to defeat the other]. ==> The candidates expected [PRO to defeat each other]. o otherwise with that if Aux is Tense, b. The men each expected [the soldier to shoot the other]. =X=> *The candidates expected [the soldier to defeat each other]. o otherwise with for. Normal proposal: The rules are language particular, the conditions on functioning universal. Tantalizing Possibility: The rules are universal (or some of them, anyway), and the conditions on functioning are language-specific? This is a key step towards Principles and Parameters, but then he retreats, but most of the paper in fact takes up arguably universal conditions on the functioning of rules.] (9) Specified Subject Condition: version 1 No rule can involve X, Y in the structure . . .X . . .[α . . . Z . . . -WYV . . . ] where Z is the specified subject of WYV in α. [26] -2Specified subject: A subject NP that contains either lexical items or a non-anaphoric pronoun. "Within the extended standard theory, as developed in the references cited earlier, both NP and S are nodes to which cyclic operations apply, and the notion "subject of" is defined not only in S, but also in such NPs as [(10)], where John, in all cases, is the 'subject,' in an extended sense of this term: (10) a. John's refusal to leave b. John's picture of Bill c. John's strategy for victory. Correspondingly, in [(9)] α can be either NP or S." [p. 91, emphasis mine] NOTE ==> by "S", he means "CP" (S-bar) [see below] ƒ ....Disjoint Reference Rule (Principle B): a rule of interpretation (RI) that marks as disjoint two NPs in the configuration [NP – V – NP – X]. (15) Unlike Person Constraint case a. I expected them to visit me. b. *I expected me to visit them. (16) Disjoint Reference (Postal 1969) a. We expect them to visit me. b. *We expect me to visit them. ƒ ..not-many rule: (17c), if good, means "J's pix of many of the children are such that I didn't see them". So another RI obeys the SSC. [p. 95] (17) Assumption: *PRO each. Because "subject" is defined as the specifier of a "cyclic node" (NP or S)... ƒ ...each-movement also fails to apply downward into an NP with a subject: (11) ...and wh-movement fails to apply (upward) from an NP with a subject: (12) ƒ Likewise (Lasnik, p.c. to NC) for not-enough association, where the judgment concerns actual grammaticality (18) a. The men each saw pictures of the other ==> The men saw pictures of each other. b. The men each saw John's pictures of the other =X=> *The men saw John's pictures of each other. ƒ ƒ a. Who did you see pictures of __? b. *Who did you see John's pictures of __? [p. 91] ..It-replacement (Tough Movement) applies out of a clause with a PRO subject, not out of a clause with a lexical NP: (13) a. The hard work is pleasant [for the rich] [PRO to do]. b. *The hard work is pleasant [for the rich] [for poor immigrants to do]. (14) a. Latin is a waste of time [for us] [PRO to learn]. b. *Latin is a waste of time [for us] [for them to teach us]. a. I didn't see many of the pictures. b. I didn't see pictures of many of the children. c. I didn't see John's pictures of many of the children. a. You didn't understand the proofs of enough of the theorems for me to be justified in giving you an A. b. *You didn't understand Euclid's proofs of enough of the theorems for me to be justified in giving you an A. 3. The Strict Cycle / The COMP escape hatch If the Strict Cycle Condition in (19) holds, wh-movement must be a cyclic rule, since it applies in indirect questions and relatives: (19) Strict Cycle Condition [(51)] No rule can apply to a domain dominated by a cyclic node A in such a way as to affect solely a proper subdomain of A dominated by a node B which is also a cyclic node. Note: The Strict Cycle Condition is introduced here for a curious purpose -- to provide a general basis for believing a particular analysis (successive-cyclic wh-movement) to be true. At the end of the section, it does some real work. -3(20) Notations to watch out for! S —> COMP S' S' —> NP Aux VP (24) [(54) -- NOTE S' and S have opposite senses from the tradition!] (21) Phase Impenetrability Condition TSC/SSC(revised) + COMP Escape Hatch No rule can involve X, Y in the structure . . .X . . .[α . . . Z . . . —WYV . . . ] where (a) Z is the specified subject of WYV in α, or (b) Y is in COMP and X is not in COMP, or (c) Y is not in COMP and α is a tensed S, or [p. 98, (55)] Remember: ƒ Tensed S means Tensed S'! ƒ α ranges over the phases cyclic nodes NP and "S" ƒ Thus movement to COMP within a single "S" is ok, even if it crosses a subject. ƒ Example (12b) is ruled out because NP is a node with a "subject" but has no COMP position, so movement crosses both a subject and a cyclic node. WH-island condition for tensed wh-islands: ƒ Example (22) is excluded because there is only one COMP escape hatch from the TSC in the embedded question and the Strict Cycle in (19) prevents a countercyclic use of that escape hatch. Superiority Condition No rule can involve X, Y in the structure . . .X . . .[α . . . Z . . . —WYV . . . ] where the rule applies ambiguously to Z and Y and Z is superior to Y. [NB: A constraint on rule application, not a constraint on rule form. Restricts the ambiguity of application] The Superiority Condition requires matrix wh-movement to pick the wh-phrase in COMP, not the remaining in-situ phrase. Application of wh-movement to the phrase in CP leaves an interrogative COMP empty: [p. 102] A rule of interpretation [p. 153] excludes this: (25) Rule of interpretation for +WH COMP A +WH COMP is interpreted only when it contains a wh-phrase. ƒ Excludes ƒ (26) just as wh-island violations are excluded. (26) *John knows what who saw. (27) "Subjacent" If X is superior to Y in a phrase marker P, then Y is subjacent to X if there is at most one cyclic category C ≠ Y such that C contains Y and C does not contain X. [p. 102] 4. Superiority Condition/Subjacency Condition (28) Assumption: PRO is not a Specified Subject. Problem: Infinitival indirect questions are islands. Subjacency Condition [p. 103, (80)] No rule can involve X, Y -- X superior to Y -- if Y is not subjacent to X. (29) a. COMP he believes [SCOMP John saw who] (22) (23) *What did he wonder where John put ___ ___. [(57)] a. %What crimes does the FBI know how to solve. b. *What crimes does the FBI know whether to solve. c. *What books does John know to whom to give. d. *To whom does John know what books to give. [ok wh-movement] b. COMP he believe [NP the claim [S COMP John saw who]] [*wh-movement] -4- 5. Subjacency extensions: controversial judgments (30) (31) [*]Who did you hear stories about a picture of __. [Chomsky] vs. [ok]What books does the government prescribe the height of the lettering on __. [Ross 1967] A problem left open. ƒ Likewise for the RI of Disjoint Reference: (37) We expected pictures of me to be on sale. ƒ "Proper containment" matters. It is ok to move each to a subject: (38) [ok] They expect each other to win. [p. 109. (104a)] 7. No raising to object ƒ 6. Extraction out of subjects (32) *Who did [stories about __ ] terrify John? (33) *What did [NP [S that John saw __]] surprise Mary? (34) Subject condition [version 1] In structures of the general form . . .X . . .[α . . . Z . . . —WYV . . . ] α may not be a subject phrase properly containing Y. ƒ Creates a redundancy with the Subjacency condition where α is an S contained in an NP, which includes sentential subjects (for Chomsky) as well as CNPC cases.[p. 108] ƒ Eliminate the redundancy: restrict the Subject Condition to cases in which Y is subjacent to X. This has no empirical effect on wh movement, but ... ƒ ...does have an effect on each-movement. Condition (34) unmodified incorrectly excludes each-movement applying across S and NP in (35): (35) We expect [S [NP pictures of each other] to be on sale]. (36) Subject condition [version 2] No rule can involve X, Y in the structure . . .X . . .[α . . . Y . . . ] where (a) α is a subject phrase properly containing MMC(Y) (i.e. the smallest maximal projection containing Y) and (b) Y is subjacent to X. So not being subjacent in this case can make things good! Pretty tricky, this guy. "There is no necessity for a rule raising the subject of an embedded sentence to the object position of the matrix sentence." [p. 113] ==> NB This "economy" argument still presupposes a transformational rule system. ƒ Perhaps we should forbid string-vacuous movement. [pp. 113-114] 8. "Specified subject" replaced by reference to a subject's controller (39) *We persuaded Bill [PRO to kill each other]. [Will Leben, p.c. to NC, p. 114 (112)] (40) We persuaded Bill [PRO to kill us]. ƒ vs. promise (41) Everything but the Kitchen Sink No rule can involve X, Y (X superior to Y) in the structure . . . X . . .[α . . . Z . . . —WYV . . . ] . . . where (a) Z is the subject of WYV and is not controlled by a category containing X, or (b) α is a subject phrase properly containing MMC(Y) and Y is subjacent to X, or (c) Y is in COMP and X is not in COMP, or (d) Y is not in COMP and α is a tensed S. [Note the striking unwillingness to allow PRO to function like any other NP!] -5- 9. Other cases where controller is relevant ƒ Arbitrary null subject (∆) blocks each-movement and RI: (42) *We heard about plans to kill each other. (43) ƒ a. They will obey [any request to kill each other] b. *They will okay [any request to kill each other] But note the still puzzling "honey-dripping" problem on p. 124! (44) a. Why are John and Mary letting the honey drip on each other's feet? b. *Why are John and Mary letting Bill drip honey on each other's feet? c. Why are they letting the baby fall on each other's laps? d. *Why are they letting Bill drop the baby on each other's laps? [Bordelois, p.c. to NC; (a) from Kayne] "The examples suggest that a notion of "agency" is involved and that perhaps the notion "specified agent" is the critical one rather than formal subject." 10. First use of the word "trace" Why does an overt subject block Tough-movement, but PRO does not? Suggestion: ƒ The moved NP actually replaces PRO as the first step of movement. ƒ Reasonable, since the NP replaces it as its final step. ƒ So far so good -- -but this story messes up on (45c), however, on the assumption that the kids is not the downstairs subject. Toys should be in the way if it moves through the PRO position. Solution [?!]: ƒ "We might account for [(45b)] by assuming that when the NP John replaces it...it leaves behind a "trace" which it controls. The trace might be PRO, or it might be the null element if we think of the transformation as moving only the terminal symbol John to the NP subject position of the matrix sentence, leaving unaltered the nodes that originally dominated it...If the trace is PRO, it will be deleted by a final rule that deletes controlled PRO in such sentences as We expected to leave. The controlled trace blocks each-Movement, so that [(45b)] is ungrammatical." [p. 131] ƒ We now can drop the ordering assumption about it-replacement and eachmovement. This lets in (45c) [though it's not clear to me how this works, consistent with the Strict Cycle and traces]. ƒ Something else blocks each-movement in (45a). [p. 132] ==> This somewhat confused discussion nonetheless makes clear one of the ways in which the concept of "trace" can eliminate stipulated ordering among rules. If R1 moves X, and R2 is sensitive to the pre-R1 position of X, one approach orders R2A) -- adjectives disallow NP complements: (63) a. Mary was taught French. b. Mary was untaught (*French). Towards LFG...