On some Prepositions that Look DP-internal : English of and French de

Abstract The article shows that, in spite of their DP-internal appearance, many instances of the English preposition of and of the French preposition de can be reanalyzed as being VP-external. Moreover, it is argued that what looks like movement of a bare quantifier turns out to be remnant movement. It is also claimed that each +N subpart of DP must get its own Case, which means that (non-head) phrases never have Case. Key words: sintax, noun phrase, prepositions, Case, movement. 1. Mostly English 1.1. Introduction The prepositions to be considered in this paper are primarily English of and French de (d’ before a vowel) in sentences like: (1) John has lots of money. (2) Jean a beaucoup d’argent. Jean has a.lot of.money In such examples of and de appear to be contained within the phrases lots of money and beaucoup d’argent, respectively, and similarly for: (3) John was admiring a picture of Mary. in which of appears to be contained within a constituent a picture of Mary. This last example allows: (4) Who was John admiring a picture of? *. I am grateful to Guglielmo Cinque, Viviane Déprez, Gemma Rigau and Joan Mascaró for helpful comments on a draft of this article. Catalan Journal of Linguistics 1, 2002 71-115 On some Prepositions that Look DP-internal: English of and French de* Richard S. Kayne New York University. Department of Linguistics 719 Broadway. New York NY 10003 [email protected] Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 71 Chomsky (1977) proposed (based in part on Bach and Horn (1976)) that this kind of ‘subextraction’ is possible only if a ‘readjustment’ rule has previously applied, breaking up the object phrase. In Kayne (2000, 316) I suggested that no readjustment rule is needed - if one gives up a certain standard assumption about prepositions. The proposal was that of in (3) can be merged outside the VP, in which case a picture of Mary in (3) will automatically not be a phrase/constituent. I would now like to consider the possibility that the same holds of (1) and (2), i.e. that of and de in those, too, can be merged outside VP, despite appearances. For English (1) this is clearly called for, given the proposed account of (4), since extraction of a parallel sort is sometimes possible in the case of (1), too: (5) Money John has lots of. I will begin by spelling out the proposal for English, and then turn in more detail to French. 1.2. P and K merged above VP The derivation I suggested for (4) rested on the idea that of in (3) (and (4)) can be merged outside VP: (6) …admiring [John a picture] →merger of of …of admiring [John a picture] →movement of John to Spec,of …Johniof admiring [tia picture] →merger of W and raising of of …ofj+W Johnitjadmiring [tia picture] →movement of VP to Spec,W …[admiring [tia picture]]kofj+W Johnitjtk More recently, in Kayne (2001), I adopted a suggestion of Ur Shlonsky’s (p.c.) that, transposed to this case, would amount to replacing the W above of in (6) by an Agr-of that is below of (and selected by of). The revised derivation of (3) would then look like: (7) …admiring [John a picture] →merger of Agr-of …Agr-of admiring [John a picture] →movement of John to Spec,Agr-of …JohniAgr-of admiring [tia picture] →merger of of …of JohniAgr-of admiring [tia picture] →movement of VP to Spec,of …[admiring [tia picture]]jof JohniAgr-of tj This derivation (like that of (6)) has the intended property that a picture of John (in the last line) is not a constituent. 72 CJL 1, 2002 Richard S. Kayne Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 72 A possible further improvement is suggested by the observation that Agr(eement) is the name of a relation, strictly speaking, and therefore cannot plausibly be drawn from the lexicon. Thinking of Bayer et al.’s (2001) revealing use of K(ase) in their study of German,1let us, then, replace Agr-of in (7) by K-of. (Whether this K-of (or K-de) is closer to genitive or to dative will not matter for what follows.) In German, K is often realized with overt Case morphology (more on D than on N), in particular in the presence of a preposition. An example is: (8) mit dem Mann with the+Kdative man The suffix -m is the dative Case morpheme K that here cooccurs with the preposition mit. (In English and French, K never has overt realization (other than with some pronouns).) In addition, I take K, like P, to be mergeable above VP (and to have an EPP feature). The derivation of (3) is now: (9) …admiring [John a picture] →merger of K-of …K-of admiring [John a picture] →movement of John to Spec,K-of …JohniK-of admiring [tia picture] →merger of of …of JohniK-of admiring [tia picture] →movement of VP to Spec,of 2 …[admiring [tia picture]]jof JohniK-of tj From the perspective of Chomsky (2001), K in (9) should turn out to be an interpretable head parallel to T (or Asp or v),3K should be associated with a set of phi-features, and the phrase moving to Spec,K should have (abstract) structural Case. Overt realization of the phi-features is arguably what we call adpositional agreeement.4That K is interpretable is suggested by the well-known fact that with certain locative prepositions in German the accusative vs. dative distinction correlates with a difference in interpretation (directional vs. non-). In addition to sharing with (3) the possibility of preposition-stranding, (1) shares with (3) the more specific property that the stranding of of is degraded if of is followed by a particle: On some Prepositions that Look DP-internal CJL 1, 2002 73 1. On K, see also Bittner and Hale (1996) and Siegel (1974), and, for a general perspective into which the pair P,K might be integrable, Simpson and Wu (2000). 2. In a partly different way, movement of VP to Spec,P is also found in Barbiers (1995, chapter 4). 3. A separate question is whether certain prepositions, e.g. of, are themselves uninterpretable (‘empty prepositions’ - cf. Chomsky’s (1995) Full Interpretation principle). 4. Which may require that P and K end up contiguous with one another, thereby possibly excluding adpositional agreement in SVO languages - this would differ from the account suggested in Kayne (1994, 49). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 73 (10) Tell me who you’re touching up a picture of. (11) ??Tell me who you’re touching a picture of up. This effect can be traced back to the fact that the last step of (9) will carry the particle along with the VP, resulting in (10). (11) can be derived only through recourse to something extra (particle preposing, in Kayne (2000, chapter 13)). The same effect is seen with (1): (12) Money he’s been handing out lots of. (13) ??Money he’s been handing lots of out. This parallel behavior suggests, in combination with the very existence of (12), that (1) allows a derivation comparable to the one given for (3) in (9): (14) …has [money lots] →merger of K-of …K-of has [money lots] →movement of money to Spec,K-of …moneyiK-of has [tilots] →merger of of …of moneyiK-of has [tilots] →movement of VP to Spec,of …[has [tilots]]jof moneyiK-of tj Again as intended, lots of money in the last line of (14) is not a constituent.5 1.3. of and Case theory Of is not always possible: (15) John has (*of) money. This is brought out by the familiar pair: (16) They destroyed (*of) the documents. (17) They approved the destruction *(of) the documents. 74 CJL 1, 2002 Richard S. Kayne 5. It might be that a picture of John and lots of money are sometimes constituents (when there’s no extraction); the text discussion could be adjusted accordingly. As for exactly why extraction of who or money from within a constituent of the form a picture of who or lots of money would not be possible, Chomsky’s (2001, 13) Phase-Impenetrability Condition may be relevant. For the question of (apparent) pied-piping in: (i) (?)the person of whom John was admiring a picture See Kayne (2001, Appendix). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 74 The usual way of approaching this last pair is to say that objective/accusative Case is not available in derived nominals. Thus in (17) the documents needs some other (structural) Case, which is provided by of (or by K-of). There is also the converse question of the obligatory absence of of in (15) and (16). Let me assume that of/K-of cannot be introduced above VP earlier than the head that is responsible for objective Case.6Consequently the direct objects in (15)-(16) will be checked for objective Case. That in turn will make them ineligible for further Case-related movement (‘inactive’ - Chomsky (2001, 6)), in particular for movement to Spec,K-of. Partially similar to these facts is the contrast: (18) John went to Paris. (19) *John went of Paris. Case-licensing by of is not sufficient. Arguably, this is because of cannot be merged prior to to. One might think that overt to is needed for theta-assignment, but the following suggests otherwise: (20) John went there. There would seem to be a null preposition here (cf. Emonds (1985), Larson (1987)), given: (21) John’s trip there went smoothly. If of could be merged prior to to, it could presumably be merged prior to the null counterpart of to, in which case (19) might be derivable, incorrectly. More directly important to the main thread of this paper than (18)-(21) is: (22) John bought a pound of apples. (23) *John bought a pound apples. The question is why apples here needs Case (‘provided’ by of/K-of) in addition to the objective Case associated with the containing phrase a pound of apples. I would like to pursue an approach that amounts to taking Vergnaud’s original Case proposal to the extreme. To take one simple example, he reasoned as follows. In some languages, the documents in (16) would have morphological Case. If we assume that it has abstract Case in English (and similar languages), we can account for certain restrictions (e.g. concerning the availability of overt subjects of infinitives). Consider now: On some Prepositions that Look DP-internal CJL 1, 2002 75 6. I am leaving aside the interaction of of with subjects. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 75 (24) They helped those ten important people. In a language like Russian, the demonstrative, the numeral, the adjective and the noun would each bear (suffixal) morphological Case. Assuming that each of those four is nominal (+N), we can now understand the UG Case filter in much the way that Emonds (2000, 351) does:7 (25) +N Case Filter: Every nominal (+N) element requires Case. Russian visibly displays the required Case on each of the subparts of the object, in sentences like (24). English displays it visibly on none. Other languages (e.g. German, Hungarian) might display it visibly on some, but not all of the subparts. An independent question is whether the specific Case found on each of the four subparts of an argument is the same. In Russian, it sometimes is, sometimes not.8 The important point is that each requires some Case, which may be realized in a visible fashion (or not). Given the +N Case Filter (25), it follows immediately that apples in (22)/(23) needs Case (despite not corresponding to a full argument). The English facts concerning the presence of overt of are of course different with numerals: (26) John bought three apples. (27) *John bought three of apples. This is true even if the numeral takes an indefinite article: (28) John bought a million apples. (29) *John bought a million of apples. The fact that ‘subextraction’ is possible in (22): (30) Apples he bought a pound of. indicates, given our earlier discussion, that of and K-of can be merged above VP: 76 CJL 1, 2002 Richard S. Kayne 7. Although he wasn’t directly concerned with DP-internal Case (see his p.362, note 39). If the -ing of English gerunds is +N, the difference between (25) and Chomsky’s (1981, 49, (6)) is diminished (cf. also Rouveret and Vergnaud (1980, 190)). On Case and Romance infinitives, whose integration into the text proposal remains to be accomplished, see Kayne (2000, 285) and references cited there. I am taking nominative to be a Case (contrary to Bittner and Hale (1996, 6)) whose frequent realization as zero is to be compared to that of third person and of singular in agreement systems. 8. See Franks (1995, chapter 4). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 76 (31) …bought [apples a pound] →merger of K-of …K-of bought [apples a pound] →movement of apples to Spec,K-of …applesiK-of bought [tia pound] →merger of of …of applesiK-of bought [tia pound] →movement of VP to Spec,of …[bought [tia pound]]jof applesiK-of tj Prior to what is shown in (31), the containing phrase ‘apples a pound’ has had its objective Case checked (see the discussion of (15) and (16)). The pair of/K-of (more precisely, K-of alone) is responsible for checking the Case of apples. Two questions arise. Why does apples in (28) not need of/K-of? And how does (a) pound in (22) (and (a) million in (28)) meet the Case requirement imposed by the +N Case Filter formulated as in (25)? The answer to the first of these must take into account the fact that English differs in this area from some other nearby languages. For example, the need for of seen in (22) does not hold in German (see van Riemsdijk (1998)), and the required absence of of in (28) distinguishes English from French: (32) Jean a acheté un million de pommes. Jean has bought a million of apples (33) *Jean a acheté un million pommes. I take it to be (virtually) certain that pommes in (32) is Case-licensed parallel to apples in (31).9In addition, the Case-licensing of apples in (28) is arguably identical to that of apples in (26). Both (26) and (28) are probably to be thought of as similar to the Russian counterpart of (24), in which all four parts of the object would bear the same (dative) Case. In other words, the proposal is that in (26) the numeral three and the noun apples bear the same (here, objective/accusative) Case. 1.4. Case is limited to lexical items The standard assumption is that this Case is also the Case of the whole phrase three apples. From a minimalist perspective, this is a notable redundancy, given (25). (Why should three apples need to have Case in addition to its (immediate) +N subparts having that same Case?) The alternative (which I take to be the more attractive) is to take Case to be a feature of lexical items only. In (26), three and apples will each have structural Case that will be valued under agreement with a probe (v, in Chomsky (2001, 6)). On some Prepositions that Look DP-internal CJL 1, 2002 77 9. To consider the possibility of an objective (English) vs. oblique (French) difference of the sort proposed for prepositions in Kayne (1981) would take us too far afield. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 77 Valuation (i.e. assignment of a value under agreement) will take place separately for three and for apples, though the result will look like Case-agreement, much as in Chomsky’s (2001, 18) discussion of past participles (except that here what is at issue is Case within DP). Independent principles will (at least in English) prevent three from being moved to the Spec of that probe independently of apples and vice versa. Conversely, principles of pied-piping (to be worked out) will allow three apples to move to the Spec of the relevant probe. Although apples in (26) can have its Case valued by v, that must in English not hold in (22), with a pound instead of three. On the other hand, English a million does act like three,10 whereas French un million acts like a pound. As in the discussion of (15) earlier, the assumption will be that Case (on apples) will be checked/valued by vif it can be (i.e. if apples is accessible to v), thereby precluding subsequent merger of of/K-of, so that (27) and (29) are excluded.11 In (22) (and (32)), apples is not accessible to v(presumably as the result of a blocking effect induced by a pound,12 in English as opposed to German). Subsequent merger of K-of will have the structural Case of apples valued by K-of and apples will move to Spec,K-of, as in (31). (The structural Case of a pound will, however, have been valued by vwhether movement of a pound takes place is not clear.) 1.5. English few/little; unpronounced NUMBER/AMOUNT There is a distinction in English between: (34) John has few books. 78 CJL 1, 2002 Richard S. Kayne 10. If ais nominal, the phrase a million apples will contain three elements each bearing the same Case. Possibly this is the result of DP-internal movement (see Uriagereka’s suggestion mentioned in Kayne (1994, chapter 8, note 53)), which leads to the question of: (i)*I have apples three. French simple numerals are like English: (ii) Jean a trois (*de) pommes. Jean has three of apples With right-dislocation, we have: (iii) Jean en a trois, *(de) pommes. Jean of.them has three of apples where Case-valuation of pommes by vis not available, presumably because right-dislocation involves (leftward - see Cecchetto (1999) and Villalba (1999)) movement. Right-dislocation with no (overt) clitic acts the same: (iv) Jean a pris les rouges, *(de) pommes. Jean has taken the red of apples 11. Beyond the scope of this article is English ‘s, as in a woman’s picture vs. *butter’s pound. I also leave aside here the question of nominative Case. 12. In which case, the relative order of a pound and apples within the DP in the derivation (31) should be reversed, and similarly elsewhere. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 78 (35) John has little money. Little cannot be replaced by few: (36) *John has few money. Nor, keeping the interpretation constant, can few be replaced by little: (37) *John has little books. This kind of number agreement is not found in French, where peu is used both with plurals and with singulars: (38) Jean a peu de livres. Jean has few of books (39) Jean a peu d’argent. Jean has little of.money The fact that peu in these examples must be followed by de: (40) *Jean a peu livres/argent. will play a role later on. For now, it is sufficient to note that we can express this by saying that peu blocks the (accusative) Case valuation of livres/argent that holds in (26) and (28), as well as in (34) and (35). The few/little distinction is of course paralleled by:13 (41) John doesn’t have many/*much books. (42) John doesn’t have much/*many money. English has no of in such cases: (43) *John has few of books. (44) *John doesn’t have many of books. On some Prepositions that Look DP-internal CJL 1, 2002 79 13. In my English, simple much is a polarity item (cf. Klima (1964, 284)): (i)*John has (very) much money. To a lesser extent, this holds for me with many: (ii) John doesn’t go to many concerts. (iii) ?John goes to many concerts. (i) contrasts with: (iv) John has much more money than Bill. Also: (v) John is smarter than Bill by a lot/*much. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 79 French also has a counterpart of sufficient(ly): (97) Jean dépense suffisamment peu d’argent. Jean spends sufficiently little of.money Unlike English, French allows: (98) Jean dépense suffisamment d’argent. Jean spends sufficiently of.money The interpretation is that of: (99) (?)John spends sufficiently much money. Although (99) is not entirely natural, the word-for-word equivalent of (98) is far worse: (100) *John spends sufficiently money. As discussed by Doetjes (1997, 102) (see also Grevisse (1993, §607)), (98) is part of a more general property of French, which allows various adverbs to ‘look like’ quantifiers, e.g.: (101) Enormément d’argent a été dépensé l’année dernière. enormously of.money has been spent the.year last The obvious proposal, from the perspective developed so far, is that these adverbs are not quantifiers themselves, but rather in (98) and (101) are modifiying the French counterpart of MUCH/MANY:25 (102) …suffisamment MUCH d’argent 86 CJL 1, 2002 Richard S. Kayne 25. Cf. Selkirk (1977, note 13). Arguably, French has no exact overt counterpart of much/many at all. A candidate is tant, with complications that would take us too far afield. A pair that fits into the text discussion is: (i) Jean a trop peu d’argent. Jean has too little of.money (ii) Jean a trop d’argent. where (ii) has the interpretation ‘too much’ and the analysis: (iii) …trop MUCH AMOUNT d’argent The fact that (ii) contains MUCH and cannot contain LITTLE (i.e. the interpretation cannot be that of (i)), which is also true of (83) and (91), presumably reflects some notion of ‘markedness’ that must be flexible enough to allow plural MANY to be unpronounced in (iv) (and in (83) and (91)): (iv) Jean a trop d’amis. Jean has too of.friends Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 86 1.7. More on Case The question now is why English prohibits (100), and similarly for: (103) ?He has invited enormously many people. (104) *He has invited enormously people. (105) *He has invited enormously of people. Let me try to formulate an answer in Case terms. In (98)/(102), whose fuller structure is: (106) …suffisamment MUCH AMOUNT d’argent the noun argent is Case-licensed via de/K-de. The fact that English disallows (105) is almost certainly the same fact as: (107) *He has invited enormously many of people. English many and MANY (more exactly, ‘many NUMBER’ and ‘MANY NUMBER’) do not permit of in (on (109), see (91)): (108) *He has invited many of people. (109) *He has invited enough of people. In all of (105)-(109) NUMBER does not block (accusative) Case valuation of people by v, so that people is ‘frozen’ relative to further Case-related movement (to Spec,K-of). As for (104), a relevant fact is that French shows a sharp contrast between (98)/(101) and:26 (110) *Jean dépense très d’argent. Jean spends very of.money A possible proposal is that adverbs in -ment are (or can be) +N in French, but très is not, and that (110) is excluded as a result of: On some Prepositions that Look DP-internal CJL 1, 2002 87 26. Possible, without de, is: (i) Jean a très faim. Jean has very hunger presumably without MUCH, too. This requires further work. I also leave open the question of derived nominals: (ii) They approved the destruction *(of) the city See also Giusti and Leko (1995). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 87 (111) The structure in (106) is well-formed only if the adverb is +N. This in turn might follow from: (112) Unpronounced MUCH/MANY must be licensed by an overt +N element. ((112) assumes that enough in (88) is +N, that the licenser in (80) is +N number and that trop in (190) below is +N.) Given (112), (104) would be excluded if English -ly adverbs (arbitrarily) could not be +N. A more attractive alternative would be to look to the interaction of MUCH/MANY and agreement and to say, thinking ahead, that (106) is (for reasons to be discovered) incompatible with DPinternal phi-feature agreement of the sort discussed in (124)-(161) below (which French alone would lack).27 In extending the requirement of Case to (nominal) subparts of arguments, the +N Case Filter may allow us to integrate: (113) John bought too big a house. (114) John bought too big of a house. In (113), big and house both have their accusative Case valued by v. In the colloquial (114), this is not possible (perhaps due to a blocking effect of the adjective), so that K-of is needed to value Case on house (and perhaps also on a). If off is nominal (and can receive some sentential Case), this might carry over to the similar pair:28 (115) John fell off the table. (116) John fell off of the table. although not to: (117) John is such (*of) an idiot! (118) What (*of) an idiot! The presence of the adjective in (114) seems to be part of the answer (it must be involved in the blocking), or perhaps a good part of the answer (if off is an adjective or at least +N). 88 CJL 1, 2002 Richard S. Kayne 27. This might allow an account of the fact that Catalan lacks (98)/(101) (despite having de in some cases parallel to (97)) - given that Catalan (sometimes) has agreement in the presence of de - see Martí Girbau (2001). 28. The significance of this similarity is supported by the fact that (114) is limited to American English (according to Kennedy and Merchant (2000, 125)), as is (116) (according to Merat (1974, 229)). Merat (1974, 212, 229) also gives all of the students as less frequent in British English, raising the possibility that the blocking of simultaneous Case-valuation of the students and all is responsible for that of, too. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 88 Note that extraction is possible in: (119) What table did he fall off of? implying, given our earlier discussion, that this of can be merged outside VP. Extraction is not possible in: (120) *What did he buy too big of? This, however, may be due to the pre-N determiner in (114) (limited to asee Bennis et al. (1998)),29 in which case the of of (114) could be VP-external, too. The +N Case Filter requires that all adjectives have Case (assuming they are all +N), and thereby makes it less surprising that some are preceded by a (Caselicensing, in conjunction with K) preposition.30 French has, for example: (121) Jean a quelque chose *(de) lourd. Jean has some thing of heavy (122) Jean en a un (de) rouge. Jean of.them has one of red To judge by: (123) John has something heavy English allows accusative valuation to reach the adjective in this context, while French does not (for reasons that need elucidation). (122) may reflect two distinct structures, one like (121), one like (123).31 1.8. Phi-feature agreement Unlike English (and French), Italian shows number and gender agreement in: (124) Gianni ha poco tempo. Gianni has little(m.sg.) time On some Prepositions that Look DP-internal CJL 1, 2002 89 29. Note that, if anything, the following is less bad than (120): (i) *What did he buy too big of a? 30. On the other hand, I have no examples of demonstratives or numerals Case-licensed prepositionally. That numerals (except one) are nouns is argued by Jackendoff (1977, 128ff.). (For my purposes, +N is sufficient.) That (121) reflects (genitive-like) Case was suggested by Doetjes (1997, 155n). 31. Catalan from this perspective entirely disallows simultaneous Case-valuation of (its counterparts of) un and rouge in (122) and requires de - see Martí (1995). Adjectives in predicate position must be Case-licensed, too, given (25), as in Emonds (2000); for a different view, see Pereltsvaig (2001). The Case-licensing requirement on adjectives might allow rethinking Baker and Stewart (1997) in Case terms (as opposed to their theta approach). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 89 (125) Gianni ha poca speranza. Gianni has little(f.sg.) hope (126) Gianni ha pochi libri. Gianni has few(m.pl.) books (127) Gianni ha poche idee. Gianni has few(f.pl.) ideas Italian makes no distinction of the few/little type, but the Italian word for few/little agrees in number and gender with the noun. This is so despite the fact that the structure of, e.g., (127) is, if we transpose our earlier proposal from English to Italian: (128) …poche NUMBER idee The question is why poche agrees with idee if it’s really modifying NUMBER. Let me approach this question through English and then French. Consider: (129) John isn’t that smart. From a DP perspective, it would appear plausible to take that here to be a head whose complement is AP. On the other hand, for the case of ordinary demonstrative that occurring with NP, Giusti (1994, 249), Sigler (1997, p.106), Bernstein (1997) and Franks (1995, 101) have proposed that that should be considered to raise from a lower position into Spec,D. That kind of analysis for (129) could be given the following form. (129) is derived from a structure resembling: (130) John isn’t as smart as that. Or, thinking of Bennis et al. (1998), from: (131) ?John isn’t smart like that. This would bring together the alls of the following: (132) John isn’t all that smart. (133) John isn’t as smart as all that. Similarly for the abouts of: (134) John is about that tall. 90 CJL 1, 2002 Richard S. Kayne Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 90 (135) John is about as tall as that. Whereas the following are now seen to display a single restriction: (136) *John is well over that tall. (137) *John is as tall as well over that. These two contrast with: (138) John is well over six feet tall. Assume, then, that (132) is derived from: (139) …smart LIKE all that (or perhaps from ‘…AS smart AS all that’) via movement of all that.32 French doesn’t have that (or this) preceding adjectives, but it does have (with an interpretation close to very): (140) Jean est tout petit. Jean is all small A plausible source, parallel to (139), is: (141) Jean est petit comme tout. Jean is small like all/anything More exactly, (140) would be: (142) …petit COMME tout with movement of tout (or perhaps of ‘COMME tout’) to some higher Spec. Of interest is the fact that tout in (140) agrees in gender:33 (143) Marie est toute petite. Marie is all(f.sg.) small(f.sg.) On some Prepositions that Look DP-internal CJL 1, 2002 91 32. Bowers’s (1975, 540) proposal to relate (i) and (ii) by rightward movement: (i) John is far more intelligent than Bill. (ii) John is more intelligent by far than Bill. can be recast in terms of leftward movement of far, with (i) being: (iii) …more intelligent BY far… and similarly for two feet higher (from ‘…higher BY two feet’), etc. 33. Though arguably not in number - see Kayne (1975, §1.5) - which I take to be orthogonal to the main point. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 91 Without movement, it would not: (144) Marie est petite comme tout/*toute. Although it might not be impossible to integrate this agreement of tout/toute into a lexicalist checking approach of the sort considered in Chomsky (1995, 239),34 another possibility, thinking in particular of Bernstein (1991) on DP-internal agreeement in Walloon, would be to take (143) to be: (145) …tout -e petite with the first -e (and perhaps also the second) an independent head. If so, then (128) might be reinterpreted as: (146) pochNUMBER -e idee with -e needing a +N host at PF. A more ‘extreme’ case of this kind of syntactic dissociation of an agreement suffix (see also Julien (2000) and Koopman and Szabolcsi (2000, 39)) comes from Italian tropp- (‘too’). Consider: (147) Gianni è troppo intelligente. Gianni is too intelligent (148) Gianni è troppo poco intelligente. Gianni is too little intelligent In the spirit of preceding proposals, we have for the first: (149) …troppo MUCH intelligente Similarly, given: (150) John has too many books. (151) Gianni ha troppi libri. the natural proposal is: 92 CJL 1, 2002 Richard S. Kayne 34. Note that the lexicalist approach to agreement considered by Chomsky does not directly extend to Romance subject clitics, which are strongly agreement-like in certain respects, especially in those French and Italian dialects where subject DPs are obligatorily doubled by a subject clitic, yet are not in any obvious sense analyzable as a verbal affix. Similar questions arise for obligatory object clitic doubling. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 92 (152) …troppi MANY libri or more exactly: (153) …too many NUMBER books (154) …troppi MANY NUMBER libri Compared to English, Italian has one extra element that is unpronounced. Yet troppi (m.pl.) agrees with libri. (The other forms would be troppo, troppa, troppe.) The same holds if in place of covert MANY in (154) we have the overt Italian counterpart of few (which also agrees - cf. (126)): (155) Gianni ha troppi pochi libri. Gianni has too few books The agreeing form troppi is all the more striking as it does not appear in the corresponding predicate sentence (although pochi continues to agree):35 (156) I libri di Gianni sono troppo pochi. the books of Gianni are too few (in number) The contrast between the agreeing form troppi in (155) and the non-agreeing form troppo in (156) recalls (in part) German, in which prenominal adjectives have an intricate agreement paradigm but predicate adjectives don’t agree at all. A proposal in the spirit of (146) would be:36 (157) tropp- [ -i [ pochNUMBER [ -i libri ] ] ] with multiple number/gender heads dependent in determiner-like fashion on the presence of NP (see Kester (1996) on Germanic). (146) and (157) recall Corver (1997, 140) on the Dutch counterpart of the biggest possible N, in which the agreement suffix follows possible rather than biggest, in effect then simply following a certain complex phrase. If we now combine (154) and (157), (151) will be: (158) tropp- [ [ MANY NUMBER [ -i libri ] ] ] whose wh counterpart (how many books) will be: On some Prepositions that Look DP-internal CJL 1, 2002 93 35. Franca Ferrari (p.c.) points out that clitic ne also yields non-agreement of tropp-: (i) Gianni ne ha troppo pochi. Gianni of.them has too few 36. Whether tropp-/too could reach their surface position by movement as was suggested for (143) and (129) is left an open question. Note that Spanish demasiado (‘too’) looks phrasal. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 93 (159) quant- [ [ MANY NUMBER [ -i libri ] ] ] in which quant-+-i is not a constituent, from which follows the fact that it cannot be moved: (160) *Quanti hai comprato libri? how-many have-you bought books The impossibility of the corresponding Engish sentence: (161) *How many have you bought books? would follow in the same way if English has an unpronounced counterpart of the -i of (159) that cannot be stranded.37 1.9. a little and a lot In some contexts, they look parallel: (162) You should help them a little. (163) You should help them a lot. My proposal for little has been that it is an adjective that can modify unpronounced AMOUNT: (164) …(a) little AMOUNT (money) Lot, on the other hand, appears to itself be a noun (that does not modify AMOUNT or NUMBER), as reflected by the following contrasts: (165) John has little/*lot money. (166) John has a little/*a lot money. (167) John has a lot of/*a little of money. (168) John has a whole/an awful lot of/*a whole/*an awful little (of) money. The of of these last two can be VP-external, given: 94 CJL 1, 2002 Richard S. Kayne 37. Similarly for Italian quali libri (‘which books’): (i) qual- [ -i libri ] and for English which books. In this vein, note the contrast: (ii) **Whose were you talking to sister? (iii) ??Who were you talking to ‘s sister? See Kayne (2000, 109). On *un gran numero -e (di) tavole, cf. in part enough in Dutch - van Riemsdijk (1992, 507). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 94 (169) (?)What (else) does he have a lot of? The derivation will be: (170) …has [what a lot] →merger of K-of …K-of has [what a lot] →movement of what to Spec,K-of …whati K-of has [ti a lot] →merger of of …of whati K-of has [ti a lot] →movement of VP to Spec,of …[has [ti a lot]]j of whati K-of tj →wh-movement A lot of differs from a little: (171) *What (else) does he have a little? That the key is of is further suggested by: (172) (?)What (else) does he have hundreds of? (173) *What (else) does he have a hundred? The reason must be as follows. In agreement with Chomsky (1977, 114), successful ‘subextraction’ of the sort found in (169) and (172) depends on a lot of what, hundreds of what not being a constituent (when wh-movement takes place). My proposal has been that the required non-constituency comes about, as seen in (170), through Case-related movement (here, of what) to Spec,K-of. If in the absence of overt of, no such Case-related movement can take place, the impossibility of (171) and (173) will follow.38 On some Prepositions that Look DP-internal CJL 1, 2002 95 38. This suggests that English has no covert of capable of inducing movement in that way, and that accusative Case-valuation of the noun of the sort seen in (165) and (166) with little, as well as in: (i) He has a hundred books. takes place without movement. The reason for the contrast: (ii) *He has a hundred of books. (iii) ?What (else) does he have a hundred of? probably involves the relation between (i) and: (iv) He has a hundred of those books. which is beyond the scope of this article. Other factors than just the presence of of must come into play with extraction, since (for me) extraction is not possible in: (ii) How big of a house did they buy? (iii) *How big did they buy of a house? On the other hand, Henry (2001) gives for Belfast English: (iv) How good has there seemed of a guide to be showing people around? Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 95 From a subextraction perspective, on the other hand, all that is preposed in (202) is the bare quantifier tant (i.e. ‘tant NUMBER’), which has no phi-features of its own, and in fact in French can never even display any via agreement: (203) *Il a tantes repeintes de bagnoles… (204) *Il a repeintes tantes de bagnoles… The (marginal) past participle agreement seen in (202) would thus be harder to understand than under the remnant movement hypothesis. 2.5. The blocking effect of prepositions A remnant movement approach to (200) that is based on a VP-external de differs from a subextraction approach in the way it interprets the blocking effect of prepositions, illustrated in: (205) Jean a souri à peu d’enfants. Jean has smiled at few of.children (206) *Jean a peu souri à d’enfants. In Kayne (1981) I had accounted for the contrast between (190) and (206) in ECP terms, taking there to be an empty QP in both, with that empty QP properly governed by the verb in (190) but not properly governed by either the verb or the preposition in (206). From the present perspective, the question is why the following derivation, which tracks (199)/(200), does not result in an acceptable sentence: (207) …souri à [enfants peu] →merger of K-de …K-de souri à [enfants peu] →movement of enfants to Spec,K-de …enfantsi K-de souri à [ti peu] →merger of de …de enfantsi K-de souri à [ti peu] →movement of VP to Spec,de …[souri à [ti peu]]j de enfantsi K-de tj →movement of ‘[ti peu]’ out of VP to the left …[ti peu]k [souri à tk]j de enfantsi K-de tj 102 CJL 1, 2002 Richard S. Kayne The less than perfect status of (202) for Déprez might be related to the indefiniteness of the preposed object. Some speakers who have past participle agreement reject (202). There might be a correlation with whether or not they accept past participle agreement with clitic en. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 102 I think that the answer lies in the last step, which arguably constitutes a preposition stranding violation of exactly the sort found in: (208) *Qui as-tu souri à? who have.you smiled at In (208) qui has been moved away from à; in (206)/(207) it is ‘[tipeu]’ that has been moved away from àin parallel fashion, with a equally unacceptable result in both cases.49 Qui and peu (more exactly ‘[tipeu NUMBER]’) differ, on the other hand, in that the pied-piping counterpart of (208) is straightforwardly acceptable: (209) A qui as-tu souri? whereas the pied-piping counterpart of (206)/(207) is, for almost all speakers, not:50 (210) *Jean a à peu souri d’enfants. This is not a property specific to …peu…de…, to judge by the following: (211) *Jean n’a compris rien. Jean NEG.has understood nothing (212) Jean n’a rien compris. (213) Jean n’a pensé à rien. Jean NEG.has thought to(of) nothing (214) *Jean n’a à rien pensé.51 Non-prepositional object rien moves leftward obligatorily (in the absence of modification), for all speakers.52 Comparable movement of prepositional object rien is generally rejected. However, I have come across one speaker (Léna Baunaz) who accepts (214) and also accepts at least some sentences like (210). More specifically, she finds acceptable (with a certain intonation contour, including stress on the quantifier word): (215) Il avait à trop répondu de lettres ce jour-là pour… he had to too-many answered of letters that day-there to… On some Prepositions that Look DP-internal CJL 1, 2002 103 49. Preposition stranding itself now needs to be reinterpreted in non-government terms, a task beyond the scope of this paper. 50. See, e.g., Milner (1978b, 100). 51. In literary French there are some examples resembling (214) that are fairly acceptable - see Kayne (1975, chap.1, note 81). 52. See Kayne (1975, §1.6). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 103 (216) Il avait à énormément téléphoné de copains. he had to enormously(many) telephoned of friends (217) Le comité était à très peu parvenu de compromis. the committee was to very few reached of compromises (218) Il s’était à beaucoup adressé de collègues. he REFL.was to many adressed of colleagues Although these are not as ‘easy’ for her as: (219) Il avait à rien pensé de drôle depuis un moment. he had to(of) nothing thought of funny for a while the fact that they are acceptable to her is striking. 2.6. combien It is not, however, entirely unfamiliar. Obenauer (1976) studied in detail the interrogative counterpart of (190): (220) Combien a-t-il acheté de livres? how-many has.he bought of books He noted (p.11ff.) that if combien is preceded by a preposition, the sentence is often rejected, but sometimes accepted: (221) A combien s’est-il adressé de gens heureux? to how-many REFL.is.he addressed of people happy (222) A combien a-t-elle souri de garçons? to how-many has.she smiled of boys These interrogative examples are more widely accepted than (215)-(218). (I suspect that the difference is related to the higher landing site with interrogatives.53) 104 CJL 1, 2002 Richard S. Kayne 53. There may be a (yet to be explained) parallel with (English) preposition stranding - see Kayne (2000, 39). Both subtypes are possible (to some extent) with other prepositions. Obenauer (1976, 11) gives as ‘??’ (vs. ‘?’ for (221)-(222)): (i) Sur combien faut-il pouvoir compter de personnes? on how-many is.necessary.it to.be.able to.count of persons And Léna Baunaz accepts: (ii) Il avait sur beaucoup tiré de lapins ce jour-là. he had at many shot of rabbits that day-there If there is in addition to the PP a direct object, the result in both subtypes is sharply impossible if the direct object intervenes: Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 104 The acceptability of both types (even though limited) is notable, since it looks at first glance as if they must, for those who accept them, be an instance of movement of a non-constituent, i.e. to take just two of the examples, it looks like there must be movement of à très peu in (217) out of a phrase ‘[ à très peu de compromis ]’; similarly it looks like there must be movement of à combien in (222) out of a phrase ‘[ à combien de garçons ]’. Starke (2001) has proposed, however, that sentences like (221)-(222) are not that at all, but rather instances of remnant movement. (His proposal converges with one made by Androutsopoulou (1997) for Greek sentences that I think are close to these.) From the present perspective, with de VP-external in all of (220)-(222), the derivation of (220) would look like: (223) …acheté [livres combien] →merger of K-de …K-de acheté [livres combien] →movement of livres to Spec,K-de …livresi K-de acheté [ti combien] →merger of de …de livresi K-de acheté [ti combien] →movement of VP to Spec,de …[acheté [ti combien]]j de livresi K-de tj →Wh-movement out of VP [ti combien]k…[acheté tk]j de livresi K-de tj Wh-movement moves combien together with the trace of livres.54 The derivation of (222) is now: On some Prepositions that Look DP-internal CJL 1, 2002 105 (iii) *A combien a-t-il montré son article de collègues to how-many has-he shown his article of colleagues (iv) *Il avait à beaucoup montré son article de collègues. he had to many shown his article of colleagues If de collègues precedes son article, there is some improvement in both (see Obenauer (1976, 74); also Androutsopoulou (1997, 26) on Greek), for unclear reasons. The impossibility of (iii)/(iv) recalls the fact that French ‘stylistic inversion’ can usually not produce ‘V O S’ where O is a lexical direct object - see Kayne and Pollock (2001, §16). The parallelism is supported by the observation that there is substantial improvement is the object is idiomatic: (v) A combien a-t-il rendu hommage de collègues. to how-many has.he rendered hommage of colleagues 54. Fanselow and Cavar’s (2001) approach to remnant movement has in common with this the movement of a ‘large’ constituent, but differs in that it uses selective deletion, rather than prior extraction of a subpart. Luigi Rizzi (p.c.) points out that a remnant movement analysis of these combien sentences will require rethinking the intervention and weak island effects discussed by Obenauer (1984; 1994, chapter 2), Rizzi (1990, 12ff.) and Starke (2001). Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 105 (224) …souri à [enfants combien] →merger of K-de …K-de souri à [enfants combien] →movement of enfants to Spec,K-de55 …enfantsi K-de souri à [ti combien] →merger of de …de enfantsi K-de souri à [ti combien] →movement of VP to Spec,de …[souri à [ti combien]]j de enfantsi K-de tj →Wh-movement [à [ti combien]]k…[souri tk]j de enfantsi K-de tj56 In the last step, à combien is pied-piped along with the trace of enfants.57 This remnant movement derivation does not require movement of a non-constituent. 106 CJL 1, 2002 Richard S. Kayne 55. Note that this must not count as an instance of preposition stranding. Why this type of sentence is not equally acceptable to all speakers remains to be understood. (There might be a link to the judgments Milner (1978b, 156n) reports for right-dislocation.) What is moved in this step in (224) is the NP enfants. This distinguishes (222) from: (i) *De quel livre a-t-elle souri à l’auteur? of what book has-she smiled at the-author on which, see Kayne (1975, §2.8, §2.10). 56. If Wh-movement moved ‘[ti combien]’ in the last step without moving à, we would have a preposition stranding violation exactly as in (206) (and (208)): (i) *Combien a-t-elle souri à d’enfants? how-many has.she smiled at of.children 57. (224) contains an oversimplification, thinking of Kayne (2001), where àitself is argued to be a VP-external probe. A fuller derivation would be: (i) …souri [enfants combien] →merger of K-à …K-àsouri [enfants combien] →movement to Spec,K-à …[enfants combien]i K-àsouri ti →merger of à …à [enfants combien]i K-àsouri ti →movement of VP to Spec,à …[souri ti]j à [enfants combien]i K-àtj At this point K-de comes in: (ii) …K-de [souri ti]j à [enfants combien]i K-àtj →movement of enfants to Spec,K-de …enfantsk K-de [souri ti]j à [tk combien]i K-àtj →merger of de …de enfantsk K-de [souri ti]j à [tk combien]i K-àtj →movement to Spec,de …[[souri ti]j à [tk combien]i K-àtj]l de enfantsk K-de tl At this point, Wh-movement pied-pipes the phrase ‘[à [tk combien]i K-àtj]’. Two points worth mentioning are: First, the phrase moved to Spec,de is headed by à, i.e. is not a VP. Second, the phrase pied-piped by Wh-movement is not a maximal projection unless ‘[souri ti]’ has moved to some higher Spec in the transition from (i) to (ii) (Koopman and Szabolcsi (2000) may be relevant). An alternative would be to reinterpret prepositional pied-piping in terms of a high merger of the preposition - see the appendix to Kayne (2001). That would mean a derivation for (222) like: Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 106 Starke’s point about remnant movement clearly carries over to (215)-(218), e.g. the derivation of (217) would be (with de VP-external): (225) …parvenu à [compromis très peu] →merger of K-de …K-de parvenu à [compromis très peu] →movement of compromis to Spec,K-de …compromisi K-de parvenu à [ti très peu] →merger of de …de compromisi K-de parvenu à [ti très peu] →movement of VP to Spec,de …[parvenu à [ti très peu]]j de compromisi K-de tj → [à [ti très peu]]k…[parvenu tk]j de compromisi K-de tj The ‘splitting’ of the PP seen in (215)-(218) and (221)-(222) is, given the derivations (224) and (225), essentially due to the movement (out of the PP) of the NP enfants/compromis to Spec,K-de. If we ask what motivates this NP movement, the answer, from the perspective of the analysis being pursued, is Case. French combien and très peu, etc. block Case-valuation from applying simultaneously to them and to their sister constituent. Therefore K-de must appear; the movement of the NP to Spec,K-de is essential if the sentence is not to violate Case requirements. Since that Spec,K-de is VP-external, we get the effect of splitting. In Italian, the counterparts of combien and peu do not take a preposition before the NP: (226) Quanti (*di) libri hai comprato? how-many of books have-you bought (227) Hai comprato pochi (*di) libri. you-have bought few of books On some Prepositions that Look DP-internal CJL 1, 2002 107 (iii) souri [enfants combien] →merger of K-de K-de souri [enfants combien] →movement of enfants to Spec,K-de enfantsi K-de souri [ti combien] →merger of de de enfantsi K-de souri [ti combien] →movement of VP to Spec,de [souri [ti combien]]j de enfantsi K-de tj →merger of subject and aux (plus inversion) as-tu [souri [ti combien]]j de enfantsi K-de tj →merger of K-à K-àas-tu [souri [ti combien]]j de enfantsi K-de tj →movement of ‘[ti combien]’ to Spec,K-à [ti combien]k K-àas-tu [souri tk]j de enfantsi K-de tj →merger of à à [ti combien]k K-àas-tu [souri tk]j de enfantsi K-de tj Again, no recourse to non-constituent movement would be necessary. The derivation in (iii) abstracts away from questions about Wh-movement - see Pollock, Munaro and Poletto (1999) - that may be relevant to the tension between (iii) and the proposal in Kayne (2000, 292, 322) concerning the obligatory filling of functional Specs. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 107 In our terms, this means that Italian does allow simultaneous Case-valuation here, e.g. of pochi and libri in (227). In the consequent absence of K-di, the Frenchtype derivation is not available, and the (correct) expectation is therefore that splitting will not be possible in Italian: (228) *Quanti hai comprato libri? Things are a bit more complex, however, since Modern Greek allows splitting in the absence of a preposition corresponding to French de, even when what is split is a PP, much as in (221)-(222). One example from Androutsopoulou (1997, 30) is: (229) Me to BLE eghrapsa molivi (ohi me to kokkino). with the BLUE I-wrote pencil (not with the red) The derivation she proposes seems correct - the NP molivi scrambles out of the PP me to BLE molivi; subsequently the PP ‘[ me to BLE t(molivi)]’ containing the trace of molivi preposes by focus movement. The question is why Greek allows this scrambling step, but not Italian. Greek has morphological Case on lexical DPs, and Italian does not. Nor does Bulgarian, yet some Bulgarian allows (some) sentences like (229) (Steven Franks and Roumyana Pancheva (p.c.)): (230) Na TOZI sedjah stol. on THIS I-sat chair It remains unclear whether this Italian vs. Greek/(some)Bulgarian difference has to do with DP-internal syntax (transparency to extraction; or internal structure as in (159)) or with the availability of a scrambling landing site in the latter, but not in Italian.58 2.7. More on remnant movement of peu Returning to the preceding derivations, we see that the movement step indicated in the last line of (224) is simply wh-movement of a familiar sort. More interesting is the last movement step in (225) and similarly in (199)/(200). Why does peu in French move leftward, even if modified by très, and even (for Léna Baunaz) if preceded by a preposition. Part of the answer may be that the movement of peu (similarly beaucoup, trop, énormément and others) is akin to the movement of rien seen in: (231) Jean n’a rien acheté. Jean NEG.has nothing bought 108 CJL 1, 2002 Richard S. Kayne 58. Note that Italian must not have an unpronounced counterpart of French de. In addition, Catalan may show that having de is not sufficient to allow splitting of the French sort, in particular when de cooccurs with phi-feature agreement, as it can in Catalan - see Martí Girbau (2001), whose correlation between the presence of de (in molts (de) llibres (‘many of books’)) and non-specificity might be interpretable in terms of splitting vs. non-splitting. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 108 Moved rien can be modified: (232) Jean n’a absolument/presque rien acheté. Jean NEG.has absolutely/almost nothing bought and for Léna Baunaz can be preceded by a preposition, as noted earlier in (219). The movement of rien does not find a parallel with lexical negative phrases:59 (233) Jean n’a acheté aucun livre. Jean NEG.has bought no book (234) *Jean n’a aucun livre acheté. In what looks like exactly the same way, moved peu cannot be accompanied by a noun (contrasting with (190)):60 (235) *Jean a peu de livres achetés. Jean has few of books bought The commonality of this restriction is emphasized by: (236) Jean n’a rien acheté d’intéressant. Jean NEG.has nothing bought of interesting’ (237) *Jean n’a rien d’intéressant acheté. especially if (236) involves remnant movement (parallel to (199)/(200)), as in: (238) …acheté [intéressant rien] →merger of K-de …K-de acheté [intéressant rien] →movement of intéressant to Spec,K-de …intéressanti K-de acheté [ti rien] →merger of de …de intéressanti K-de acheté [ti rien] →movement of VP to Spec,de …[acheté [ti rien]]j de intéressanti K-de tj →movement of rien out of VP …[ti rien]k…[acheté tk]j de intéressanti K-de tj On some Prepositions that Look DP-internal CJL 1, 2002 109 59. Why the Icelandic counterpart of (234) is grammatical (and usual) in contrast to French remains to be understood. The French pattern recalls Cinque (1990) on bare quantifiers, though ‘bare’ must now be reinterpreted to accommodate remnant movement. 60. There may be a link here to: (i) J has carefully/??with care described it to them. (ii) J has frequently/often/?twice/*many times/*five times gone there for his vacation. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 109 It should be said that movement of rien and movement of peu differ somewhat in that movement of rien is essentially obligatory: (239) *Jean n’a acheté rien. and preferred even in cases like (236): (240) ?Jean n’a acheté rien d’intéressant. whereas non-movement of peu seems fully acceptable: (241) Jean a acheté peu de livres. Conversely, while (231) is fully acceptable without hesitation to all French speakers (that I have ever asked), movement of peu, as in: (242) Jean a peu acheté de livres. is more ‘delicate’ in a way that’s hard to pin down.61 This recalls the fact that Icelandic negative phrase movement is more robust than the movement of other quantified phrases (see Svenonius (2000)), and suggests that despite the parallelisms mentioned, the movement of rien and that of peu might not be triggered in exactly the same way, and might not, thinking of Cinque (1999), have exactly the same landing site.62 This bears in turn on a long-standing question concerning the relation between (242) and instances of adverbial peu, as in: (243) Jean a peu apprécié ta communication. Jean has little appreciated your talk I take a non-movement adverbial approach to (242) of the sort considered in Kayne (1975, §1.5) (using a non-movement mechanism that in fact mimics movement) to be undesirable (from a derivational perspective) on grounds of theoretical redundancy. In addition, the parallel with adverbs is not perfect. For example, Viviane Déprez (p.c.) finds passably acceptable: (244) (?)Elle a tout plein acheté de bouquins. she has all full bought of books but a comparable position for adverbial tout plein is very marginal for her (cf. Doetjes (1997, 178)): 110 CJL 1, 2002 Richard S. Kayne 61. See Grevisse (1993, §295) and Milner (1978b, 98). 62. Relevant is Milner (1978b, §2.3.3) on the (in)compatibility of moved peu (and similar elements) with various types of adverbs. The movement of ‘[ti peu]’ in (242) might well be expected to correlate with interpretive effects - for relevant discussion, see Obenauer (1983, 82; 1994, 111ff.), Doetjes (1997, chapter 10) and Vinet (2001, 85ff.). Some of the psych-verb restrictions discussed by Obenauer might fall out from the object being prepositional - see Landau (2001); others recall Landau (1999) on restrictions on possessive datives. Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 110 (245) ??Elle a tout plein rigolé. she has all full had-fun The remnant movement approach that I have been pursuing is compatible with these facts.63 2.8. Conclusion In conclusion, many instances of French de (‘of’) and English of that look DPinternal can be reanalyzed as being VP-external. What looks like movement of bare ‘quantifiers’ such as peu (‘few’/’little’) turns out to be remnant movement (as Androutsopoulou (1997) had proposed for Greek and Starke (2001) for interrogative combien). In many cases there is reason to postulate the presence of an unpronounced AMOUNT/NUMBER and/or an unpronounced MUCH/MANY, both in French and in English. The Case filter should apply to all +N elements, much as Emonds (2000) had suggested. In particular, each +N subpart of DP must get its own Case, sometimes in a way that parallels Chomsky’s (2001, 18) discussion of participles. This leads to the proposal that (non-head) phrases never have Case. References Androutsopoulou, A. (1997). On remnant DP-movement in modern Greek, M.A. thesis, UCLA. Bach, E. and G.M. Horn (1976). «Remarks on ‘Conditions on Transformations’». Linguistic Inquiry, 7: 265-299. Baker, M.C. and O.T. Stewart (1997). «Unaccusativity and the Adjective-Verb Distinction: Edo Evidence». Proceedings of the North East Linguistic Society, 27: 33-47. Barbiers, S. (1995). The syntax of interpretation, Doctoral dissertation, Holland Institute of Generative Linguistics. Bayer, J., M. Bader and M. Meng (2001). «Morphological Underspecification Meets Oblique Case: Syntactic and Processing Effects in German». Lingua, 111: 465-514. Bennis, H., M. den Dikken and N. Corver (1998). «Predication in Nominal Phrases». Journal of Comparative Germanic Syntax, 1: 85-117. On some Prepositions that Look DP-internal CJL 1, 2002 111 63. (Non-remnant) movement is explicitly argued for by Milner (1978a, 690-692), non-movement by Doetjes (1997, chapter 10). Whatever triggers the movement of peu (and of tout plein in (244)) must not allow: (i) *J’ai un million lu de livres. I.have a million read of books which contrasts for Isabelle de Crousaz with: (ii) Il avait un peu acheté de chocolat pour lui faire plaisir. he had a little bought of chocalate to him/her make pleasure (Obenauer (1983 p.80) had found a fairly similar example to be impossible.) Cat.Jour.Ling. 1 001-259 26/2/03 15:52 Página 111