1 On tough-movement* Milan Rezac, University ... - Multimania.co.uk

Notes
* I am grateful to comments from the audience of the Minimalist Theorizing workshop, and to Cedric Boeckx as
organizer and editor. This works develops the ideas skeched out in Rezac (2004a: chapter 3, esp. p. 187ff.), and for
comments on that section I am grateful to Gabriela Alboiu, Diane Massam, and Peter Svenonius. Mélanie Jouitteau
has patiently borne the brunt of my quest for tough-movement and copy-raising in French, and Susana Béjar, for
reconstruction therein in English. Thanks go also Glyn Hicks who pointed out a couple of inaccuracies in the nick of
time. Remaining problems are mine. This work was partly funded by SSHRC grants #752-2000-1545, #756-2004-
0389, and UPV-EHU grant #9 UPV 00114.130-160.09-2004 U.
1 This has led Chomsky (1995: chapter 3) to the reintroduction of generalized transformations, in order to allow the
S-structure insertion of a complex TM subject; cf. Uriagereka (2000), Lasnik and Uriagereka (1988:147).
2 Wilder (1991:123, 128-9) notes some intriguing but limited counter-examples involving clausal arguments:
(i) [For him to be top of the class] is hard to believe e. (Wilder 1991:123)
(ii) *I believe for him to be the top of the class. (Wilder 1991:123)
3 Keeping only to those syntactically visible at all as diagnosed by A-movement, like tabs in Tabs seem to be kept on
Kate / They keep tabs on Kate and unlike the bucket in Kate kicked the bucket / *The bucket was kicked by Kate.
4 Cf. however Lasnik and Fiengo (1974:552), who point out that the OP clause degrades as a passive infinitive, just
like the complement of restructuring verbs like try.
5 Epstein (1984) argues that it is present as a covert universal argument even if not overt. Note that there may be
other (typically left-peripheral) for-phrases around that are not experiencer arguments.
6 Indeed, Nanni (1980) points out that the OP clause cannot be pied-piped by wh-movement of the trigger, whether
or not TM occurs, (i), (ii). Thus either wh-movement takes only minimal AP projection, or the OP clause rightadjoins
higher than the AP, e.g. at the VP level, which accounts for Nanni's data without positing reanalysis.
(i) *How easy for the children to tease is John? (Nanni 1980:572)
(ii) *How difficult for them to understand the issues was it? (Nanni 1980:578)
7 Here in speaking of a DP I really mean an e-type DP. Like an e-type DP (i), a quantifier must compose with its
sister turned into a λ-abstract (type 〈e,t〉) (ii). However, while the λ-abstract in (i) takes an e-type DP as an argument,
it is the generalized quantifier denoted by a quantified DP (type 〈〈e,t〉,t〉) which takes the λ-abstract (〈e,t〉) as an
argument. The variable bound by the derived predicate's λ-abstract is interpreted through the e-type individuals that
the generalized quantifier quantifies over, e.g. most students. Though the interpretive function-argument relationship
is different in the two cases, the issues and mechanics discussed in this section are the same, so I can keep to (i).
(i) DP 〈e〉: DP λx[…x…] Pred Pred(DP) (where DP substitues for x in Pred)
(ii) [Q DP] 〈〈e,t〉,t〉: QP λx[…x…] Pred Q(DP)(Pred) (where Q binds x in Pred)
8 ⇒
However, this detail of the proposal does not play a crucial role in the treatment of tough-movement itself, which
could do with formulating Predicate Abstraction to operate on φ Agree-valued (index) features directly.
9 Cf. Branigan and MacKenzie (2001:405, 399 note 14), Rezac (2004a:136f.) for the blocking of cross-clausal
agreement by the addition of a matrix φ-accessible goal.
10 This can be implemented by giving C interveners a designated φ-set whose unique index will be compatible with
Merge of the it expletive only (cf. Lasnik 1999:136), or by assuming that the intervener is really it rather than C in a
structure [ α it CP] (Rosenbaum 1965: chapters 1, 2, Moro 1997:173ff., Anagnostopoulou 2003:187); see Rezac
(2004b) for discussion.
11 I ignore here whether interaction with phase theory needs to be further specified (cf. Branigan and MacKenzie
2001, Bruening 2001, Svenonius 2004), and whether φ-Agree is directly with the goal in violation of the Activity
Condition (op. cit.) or whether it proceeds via the head that the goal has Agreed with in its clause as in Rezac
(2004a:199ff.). Section 4 here argues that OP in tough-movement is just a C that has Agreed with the gap, avoiding
the latter question for English.
12 The glosses are taken from the cited originals and employ the following abbreviations: arabic numerals (1, 2, etc.)
for person, roman numerals (I, II, etc.) for noun class, F feminine, SG singular, PL plural, NOM nominative, ERG
ergative, ABS absolutive, DAT dative, TOP topic, OBV the obviative of the Algonquian languages, SE Romance se-type
particle, NEG negation, NMLZ nominalizing affix, ASP aspect, PERF perfect, PRES present tense, PAST past tense, PART
participle, CONJ the conjunct order of the Algonquian languages, FV the final vowel of the Bantu verb, CA
complementizer agreement, SA subject agreement.
13 Unless the experiencer is an anaphor, in which case it arguably cannot itself control PRO.
26