1

French tough-movement revisited1

J.-MARC AUTHIER AND LISA A. REED

Abstract In this paper, we argue that predicates of the tough-class in French embed not a verbal infinitive but rather, a gerundive verbal noun. This hypothesis allows us to capture a number of unexpected restrictions on French tough-movement discussed by Legendre (1986). We show that these restrictions are best described as the inability of French tough-movement infinitives to be followed by complements that are disallowed in their corresponding argument-taking event nominals. Our analysis of such infinitives as nominalized elements correctly predicts that they should never be selected by auxiliaries, and that they should have suppressed external arguments in the sense of Grimshaw (1990). French tough-movement constructions are further argued to be closely related to the type of English gerund Huddleston (1971), Hantson (1984) and Clark (1985) call ‘retroactive gerunds’. Like French tough-movement constructions, English retroactive gerunds contain a gap, are possible only with a restricted class of predicates, do not license unbounded dependencies, and do not allow subjects, except when they are realized as by-phrases.

1. Introduction It has, at times, been claimed in the literature on Romance causatives that Faire-par embeds not a verbal infinitive but rather, a gerundive, verbal noun (for example by Guasti, 1990; Travis, 1992 and Folli and Harley, 2007). In this paper, we will show that if similar assumptions are made concerning the infinitive embedded under predicates of the Tough-class in French, a number of unexpected restrictions on French Tough-movement constructions (hereafter FTMs) discussed by Legendre (1986) follow naturally. We will also establish that, given their nominal character, FTM infinitives are closely related to retroactive gerunds in English. 2. Evidence for NP-movement It has been known for some time (see e.g. Chomsky, 1982) that while English Tough-movement constructions license unbounded dependencies and parasitic gaps, as illustrated in (1a) and (1b), FTMs cannot contain more than one infinitive and do not support parasitic

1 For comments regarding an earlier version of this work, we wish to thank the audiences at the University of Kansas (fall 2007) and at the University of Pennsylvania (spring 2008). We are also indebted to two anonymous Probus reviewers for their constructive criticisms. We take full responsibility for any remaining shortcomings.

2

gaps, as can be seen in (1c) and (1d). Further, FTMs differ from French infinitival relatives in this respect as well, as illustrated in (1e) and (1f) (see Canac Marquis (1996:36-37)). (1) a. This book will be easy to tell the children to read ___. b. This panel is impossible to swing ___ without unscrewing PG first. c. *Ce livre serait difficile à empêcher ta sœur de lire ___. this book would-be difficult to to-prevent your sister of to-read ‘This book would be difficult to prevent your sister from reading.’ d. *Ce climatiseur est emmerdant à nettoyer ___ sans démonter PG d’abord. this air conditioner is annoying to to-clean without to-take-apart first ‘This AC is a bitch to clean without taking apart first.’ e. C’est un livre à être fier de dire que tu as lu ___. this-is a book to to-be proud of to-say that you have read ‘Here is a book to be proud to say that you read.’ f. Voilà bien un homme à ne pas juger ___ sans avoir rencontré PG au préalable. here-is well a man to not to-judge without to-have met previously ‘He is a man you shouldn’t judge without having met first.’ These properties of FTMs are expected under an A-movement analysis of the construction, an analysis that is, in fact, supported by a substantial body of evidence. First, passives and FTMs pattern alike with respect to object raising from VP idioms. For example, the idiomatic reading of porter assistance ‘to lend assistance’ is preserved in both constructions as can be seen in (2a) and (2b) but the idiomatic reading of casser la croûte ‘to eat’ is not as (2c) and (2d) illustrate. (2) a. Assistance sera portée aux victimes d’inondations. assistance will-be carried to-the victims of flooding ‘Help will be made available to flood victims.’ b. Assistance est difficle à porter aux victimes d’inondations. assistance is difficult to to-take to-the victims of-floods ‘Help is difficult to make available to flood victims.’ c. La croûte a été cassée par les ouvriers à midi. the crust has been broken by the workers at noon ‘The crust was broken by the workers at noon.’ *‘A meal was eaten by the workers at noon.’ d. La croûte est difficile à casser à midi (quand on est serveur). the crust is difficult to to-break at noon when one is waiter ‘The crust is difficult to break at noon (when you’re a waiter).’ *‘Lunch is difficult to eat at noon (when you’re a waiter).’ Second, Kayne (1975) observes that predicate NPs cannot undergo raising in passives as seen in (3b) and that they behave in a similar fashion in FTMs, as (3d) illustrates. (3) a. Son fils deviendra ton meilleur ami. his son will-become your best friend ‘His son will become your best friend.’

3

(3) b. *Ton meilleur ami sera devenu par son fils. your best friend will-be become by his son c. Il est facile de devenir ton meilleur ami. it is easy of to-become your best friend ‘It’s easy to become your best friend.’ d. *Ton meilleur ami est facile à devenir. your best friend is easy to to-become Third, while the impersonal French pronoun on is ambiguous between a referential first person plural reading and an existential reading when it is not a derived subject as in (4a), it can only be interpreted referentially when it undergoes object-to-subject raising in passives like (4b) and again, FTMs pattern with passives in this respect, as (4c) shows. (4) a. On te volera ta moto. ON you-DAT will-steal your bike ‘We/someone will steal your bike.’ b. On sera arrêté par la police. ON will-be arrested by the police ‘We/*someone will be arrested by the police.’ c. On sera impossible à satisfaire. ON will-be impossible to to-please ‘We/*someone will be impossible to please.’ Fourth, as noted by Kayne (1975), in FTMs, tous ‘all’ can appear in the object position of the infinitive and be understood as modifying the surface subject of the Tough-predicate as shown in (5a). And this is, once again, also possible in passives as (5b) illustrates. (5) a. Ces livres ne seront pas faciles à mettre tous dans un seul carton. these books will-be not easy to to-put all in one single box ‘These book won’t be easy to all put in a single box.’ b. Ces livres ont été mis tous dans un seul carton. these books have been put all in one single box ‘These books were all put in the same box.’ Fifth, as first pointed out in Kayne (1975), French verbs of the obéir ‘to obey’ class are exceptional in having passives in which the derived subject corresponds to a prepositional/dative complement as shown in (6a-b). Given this, it is remarkable that FTMs with obéir as their infinitive, such as the one in (7b), allow their objects to raise, just as they do in passives. (6) a. Les soldats ont obéi aux ordres/*les ordres du capitaine. the soldiers have obeyed to-the orders/*the orders of-the captain ‘The soldiers obeyed the captain’s orders.’ b. Les ordres du capitaine seront obéis. the orders of-the captain will-be obeyed ‘The captain’s orders will be obeyed.’

4

(7) a. Il est difficile d’obéir *(à) de tels ordres. it is difficult of to-obey *(to) some such orders ‘It is difficult to obey such orders.’ b. De tels ordres sont difficiles à obéir. some such orders are difficult to to-obey ‘Such orders are difficult to obey.’ Finally, as observed by Déprez (1990), NP movement in passives allows optional reconstruction in French, as shown in (8a), while wh-movement does not, as (8b) illustrates. This test, applied to FTMs by Canac Marquis (1996), again shows that they exhibit A-movement properties, as can be seen in (8c). (8) a. [Cette photo de lui-mêmei/Jeani] luii a été transmise. this picture of himself/Jean to-him has been passed-on ‘This picture of himself/Jean was sent to him.’ b. *[Quelle photo de lui-mêmei/Pauli]j est-ce que Luc luii a transmise [e]j? which photo of himself/Paul Q-particle Luke to-him has passed-on ‘Which picture of himself/Paul did Luke send to him?’ c. [Cette photo de lui-mêmei/Jeani] sera difficile à luii transmettre. this picture of himself/Jean will-be difficult to to-him to-pass-on ‘This picture of himself/Jean will be difficult to send to him.’ 3. Some problems for the NP-movement analysis Contrasts such as (9a) versus (9b), which illustrate the fact that argument nominals but not predicate nominals can undergo Tough-movement in French, suggest that the motivation for this type of object-to-subject raising is Case. (9) a. *[Ton meilleur ami]i est facile à devenir [e] i. your best friend is easy to to-become b. [Ton meilleur ami] i est facile à berner [e] i. your best friend is easy to to-con This, however, raises the question of how the (accusative) Case associated with the infinitival form of the verb berner ‘to con’ in (9b) is suspended since similar forms of the same verb retain their Case-assigning properties outside FTM, as (10) illustrates. (10) Cédric est enclin à berner ses amis. Cedric is prone to to-con his friends ‘Cedric tends to con his friends.’ A second potential problem for the NP-movement analysis of FTMs involves some troublesome breakdowns in the nice parallel that otherwise exists between verbs that can

5

passivize and infinitives that partake in Tough-movement. This problem, uncovered by Legendre (1986), is illustrated by the paradigm in (11) through (14).2 (11) a. Le camion a été chargé de tomates. the truck has been loaded of tomatoes ‘The truck was loaded with tomatoes.’ b. Il sera facile de charger le camion de tomates. it will-be easy of to-load the truck of tomatoes ‘It will be easy to load the truck with tomatoes.’ c. *Le camion sera facile à charger de tomates. the truck will-be easy to to-load of tomatoes ‘The truck will be easy to load with tomatoes.’ (12) a. Le sapin a été décoré de guirlandes. the tree has been decorated of garlands ‘The tree was decorated with garlands.’ b. Il sera facile de décorer le sapin de guirlandes. it will-be easy of to-decorate the tree of garlands ‘It will be easy to decorate the tree with garlands.’ c. *Le sapin sera facile à décorer de guirlandes. the tree will-be easy to to-decorate of garlands ‘The tree will be easy to decorate with garlands.’ (13) a. Cette grosse tringle a été enfilée de petits anneaux. this big rod has been slipped-onto of small rings ‘Someone or other slipped small rings onto this big rod.’ b. Il n’est pas facile d’enfiler cette grosse tringle de petits anneaux. it is not easy of-to-slip-onto this big rod of small rings ‘It is not easy to slip small rings onto this big rod.’ c. *Cette grosse tringle n’est pas facile à enfiler de petits anneaux. this big rod is not easy to to-slip-onto of small rings (14) a. Ta secrétaire a été chargée d’une course urgente. your secretary has been entrusted of-an errand urgent ‘Your secretary was entrusted with an urgent errand.’ b. Il sera impossible de charger ta secrétaire de cette course urgente. it will-be impossible of to-entrust your secretary of this errand urgent ‘It will be impossible to entrust your secretary with this urgent errand.’ c. *Ta secrétaire sera impossible à charger de cette course urgente. your secretary will-be impossible to to-entrust of this errand urgent

2 The judgments that appear next to the (c) examples in (11-14) are those reported by Legendre. As pointed out by our reviewers, however, there are speakers for whom such sentences are not fully ungrammatical. Nevertheless, the contrast between the (b) and the (c) examples appears to be sharp for all speakers.

6

On Legendre’s Relational Grammar account, the ungrammaticality of the (c) examples results from a violation of a constraint on Tough-movement whereby only a nominal that bears a 2-grammatical relation (i.e., direct object) on both the initial and the final strata can raise (strata being syntactic levels). For example le camion ‘the truck’ in (11c) is taken to head an initial LOCATIVE arc (due to the fact that it appears as a LOCATIVE in the related structure in (15)) and to subsequently head a final 2-arc (due to the fact that it can undergo Personal Passive, as in (11a)). This nominal is therefore ineligible for Tough-movement, hence the illicitness of (11c). (15) Il a chargé les tomates sur le camion. he has loaded the tomatoes on the truck ‘He loaded the tomatoes onto the truck.’ This account, it seems to us, is cast into doubt when one considers the grammaticality of (16) since, by the same reasoning, cette vieille péniche ‘this old barge’ also undergoes LOCATIVE to 2 Advancement and should therefore be unable to undergo Tough-movement, contrary to fact. (16) Cette vieille péniche serait impossible à charger avec du minerai this old barge would-be impossible to to-load with some ore de plomb (sans la faire couler). of lead without it to-make to-sink ‘This old barge would be impossible to load with lead ore (without sinking it).’ What has gone unnoticed until now, however, is that the Tough-movement infinitives in (11) through (14) are ungrammatical only when they are followed by those complements that are disallowed in their corresponding argument-taking event nominals, as the paradigm in (17) through (19) illustrates. (17) a. Cette péniche serait impossible à charger *de/avec du minerai de plomb. this barge would-be impossible to to-load *of/with some ore of lead ‘This barge would be impossible to load with lead ore.’ b. Le chargement de la péniche *de/avec du minerai de plomb va commencer. the loading of the barge *of/with some ore of lead goes to-begin ‘The loading of the barge with lead ore is going to begin.’ (18) a. Cette tringle est difficile à enfiler *de/avec des petits anneaux. this rod is difficult to to-slip-onto *of/with some small rings ‘It is difficult to slip small rings onto this rod.’ b. L’enfilage de cette tringle *d’/avec des anneaux prend deux minutes. the-slipping of this rod *of/with those rings takes two minutes ‘The process of slipping rings onto this rod takes two minutes.’

7

(19) a. Notre sapin sera facile à décorer *de/avec des guirlandes. our tree will-be easy to to-decorate *of/with some garlands ‘Our tree will be easy to decorate with garlands.’ b. La décoration de notre sapin *de/avec des guirlandes a pris une heure. the decoration of our tree *of/with some garlands has taken one hour ‘The decoration of our tree with garlands took one hour.’ What we intend to show is that this paradigm follows from the fact that French Tough-movement infinitives have some nominal properties. A well-known difference between active verbal forms and nominal elements is that only the argument structure of the former requires a subject. As observed by Grimshaw (1990), while complex event nominals do take obligatory objects, they never require a subject and they function, in that respect, like passive verbal forms: In both cases the argument of a passive verb or a nominal that corresponds to the external argument of the active verbal base is suppressed in the argument structure and therefore not required when it comes to satisfying argument structure in the syntax. Thus, the argument corresponding to the external argument of an active verb may occur optionally in nominals and passives as a by-phrase (or a possessive in English) as illustrated in (20). (20) a. The (enemy’s) destruction of the city (by the enemy). b. The city was destroyed (by the enemy). At least three pieces of evidence suggest that Romance Tough-movement infinitives are like nominals and passives (and unlike active verbs) in this respect. First, as observed in Montalbetti and Saito (1983), Spanish Tough-movement constructions can sometimes be morphologically identical to passive infinitivals. This is illustrated in (21). Although such Tough-passives are only found in somewhat stilted written styles (Mario Montalbetti, p.c.), the fact that they do exist establishes an intriguing parallel between Romance Tough-movement constructions and passives. (21) a. Esta vocal es facil de ser nasalizada. this vowel is easy of to-be nasalized ‘This vowel is easy to nasalize.’ b. Esta enfermedad es facil de ser curada. this sickness is easy of to-be cured ‘This illness is easy to cure.’ Second, as noted in Kayne (1975) and Canac Marquis (1996), the understood subject of French Tough-movement infinitives cannot be quantified by tous ‘all’, as shown in (22c). Since this type of modification is normally available in other infinitives, as (22a and b) illustrate, it appears that French Tough-movement infinitives do not have a canonical, syntactically realized (and phonologically null) subject, a conclusion that accords with the

8

hypothesis that they are, in fact, nominal in nature and therefore have a suppressed external argument.3

3 A reviewer asks why, if FTM infinitives do not have a syntactically realized subject, generic reflexives like soi-même ‘oneself’ and anaphoric soi ‘self’ can appear in FTMs and be understood as being referentially linked to their unexpressed subject. (i) Ce genre de claque est toujours facile à se donner à soi-même. this kind of slap is always easy to self-to-give to oneself ‘This kind of slap is always easy to give to oneself.’ (ii) Cette arme est difficile à transporter sur soi. this weapon is difficult to to-carry on self ‘This weapon is difficult to carry on one’s body.’ These facts can be viewed as problematic for our claim only if such anaphoric elements as soi-même and soi require syntactically realized antecedents. However, it is relatively easy to find examples that suggest that this assumption is unwarranted. Example (iii) shows that soi-même can take as its semantic antecedent the unexpressed external argument of a noun and example (iv) establishes that the same is true of soi. (iii) On appelle plagiat l’attribution de la propriété intellectuelle one calls plagiarism the-attribution of the property intellectual des autres à soi-même. of-the others to oneself ‘Plagiarism is the attribution of other people’s intellectual property to oneself.’ (iv) Le transport d’armes sur soi est interdit. the transport of-weapons on self is forbidden ‘Carrying concealed weapons is illegal.’ A second reviewer points out that if deverbal nouns can never have a PRO subject, the fact that control by the unexpressed agent of nominal expressions like l’incendie de voitures ‘the burning of cars’ in (v) needs to be explained, especially in view of the fact that such control fails to obtain with its equivalent (ergative) verbal counterpart, as shown in (vi). (v) L’incendie de voitures pour escroquer les assurances est devenu un sport national. the-burning of cars to to-swindle the insurances has become a sport national ‘Burning cars to collect the insurance has become a national epidemic.’ (vi) *La voiture a brûlé pour escroquer l’assurance. the car has burned to to-swindle the-insurance ‘The car (was) burned to collect the insurance.’ We believe that Williams (1985:308-313) provides a relatively solid answer to this question. According to him, the controller of the result clause in (v) is not a PRO nor is it the implicit

9

(22) a. Je dirai à ces garçons de (tous) partir (tous) en Italie. I will-say to these boys of (all) to-go (all) to Italy ‘I will tell these boys to all go to Italy.’ b. Il serait facile de (tous) contenter (tous) Paul. it would-be easy of (all) to-please (all) Paul ‘It would be easy to all please Paul.’ c. Paul serait facile à (*tous) contenter (*tous). Paul would-be easy to (*all) to-please (*all) ‘Paul would be easy (for us) to all please.’

agent in the argument structure of the noun incendie ‘fire.’ Rather, (v) is a case of event control; that is, the result clause is controlled by the N’ incendie de voitures ‘burning of cars’ in the sense that it is the circumstance ‘burning of cars’ under the control of some purposeful agent that results in fraudulent insurance claims. As he points out, this is quite different from saying that the person(s) responsible for this circumstance is/are an Agent in the theta-theoretic sense. Evidence for this view comes from the semantic oddity of sentences like (vii), which is then expected because it doesn’t make much sense to say that Marie’s arrest could indict someone. (vii) *L’arrestation de Marie pour inculper Paul a été un succés. the-arrest of Marie to indict Paul has been a success As for (vi), Williams has noted that such examples can be rendered sensible by putting them in a context in which la voiture a brûlé ‘the car burned’ is a circumstance under the control of some purposive agent; for example, if a playwright is justifying the design of a play by saying the following:

(viii) Au début de l’acte III, la voiture brûle pour permettre à l’épicier at-the start of the-act III the car burns to to-allow to the-grocer de se comporter en héro.

of self to-behave in hero ‘At the beginning of act (iii), the car burns to allow the grocer to act heroic.’

The sentence in (viii) does not mean that the playwright allows the grocer to behave like a hero; rather the fact that the car burns does. The difference between the transitive deverbal nominal incendie de voitures ‘burning of cars’ in (v) and the intransitive verb brûler ‘burn’ in (vi) then ultimately boils down to the hidden causative meaning component (i.e. ‘cause X to burn’) present in the former but absent in the latter, a component which provides the necessary controller for purposive pour ‘in order to’ clause to be felicitous. No such requirement is triggered by infinitival adjuncts introduced by sans ‘without’ as the following shows: (ix) La voiture a brûlé sans poser de risques aux habitants. the car has burned without to-pose some risks to-the residents ‘The car burned without putting the residents at risk.’

10

Finally, while French Tough-movement constructions do not easily license by-phrases, we have been able to find examples like (23), which are judged grammatical by a significant number of native speakers and show that by-phrases are not, in fact, incompatible with French Tough-movement constructions, as has often been assumed in the literature (see e.g. Kayne (1975: 337 n.73)). (23) a. Bien que ce saut soit difficile à exécuter par un débutant ... although this jump is difficult to to-execute by a beginner... ‘Although this jump is difficult for a beginner to execute...’ b. Ces abstractions ne sont faciles à visualiser que par les grands peintres. these abstractions are easy to to-visualize only by the great painters ‘These abstractions are easy only for the great masters to visualize.’ 4. FTM infinitivals as verbal nouns The hypothesis that French Tough-movement infinitivals are verbal nouns immediately provides an answer to the question of why French Tough-movement constructions have derived subjects, namely because gerundive, verbal nouns, like regular nominals, do not have Case marking abilities, hence their direct object is attracted by the nominative inducing tensed T. There is no Minimality or Minimal Link Condition violation because there is no PRO subject of the verbal noun as shown in (22c), hence we are dealing with a movement akin to that found in passives, the only difference being that passive past participles are verbal while French Tough-movement infinitivals are nominal. It thus appears that in French, what is commonly referred to as an infinitive suffix (e.g., -er, -re, -ir etc.) can sometimes be nominalizing as was proposed by Guasti (1990) in the context of Faire-par causatives in Italian. If a nominalizing suffix needs to attach to a lexical verbal root for it to acquire nominal properties then we predict that untensed perfective compounds of the form avoir ‘have’ + past participle should never occur in FTMs. This prediction is indeed correct as the paradigm in (24) illustrates. Tough-movement is impossible in (24c) because non-passive past participles like battu ‘beaten’ are verbs that bear an accusative Case feature, hence the raising of their direct object to the checking domain of tensed T is prohibited by the Last Resort Condition. That the illicitness of aspectual auxiliaries in FTMs is due to their nominal character, and not to some other factor, is further substantiated by the fact that such auxiliaries are fine in infinitival relatives, as illustrated in (24d), which we borrow from Canac Marquis (1996:41). (24) a. Aux Etats-Unis, ceci équivaut à avoir obtenu une Maîtrise. at-the United States this amounts to to-have obtained a Master’s ‘In this US, this is equivalent to having obtained a Master’s degree.’ b. Il aurait été impossible d’avoir battu ce record it would-have been impossible of to-have beaten this record sur une piste mouillée. on a track wet ‘It would have been impossible to beat this record on a wet track.’

11

(24) c. *Ce record aurait été impossible à avoir battu this record would-have been impossible to to-have beaten sur une piste mouillée. on a track wet ‘This record would have been impossible to beat on a wet track.’ d. Voilà un excellent service à avoir rendu à des amis. here-is an excellent service to to-have rendered to some friends ‘This is a great favor to have done for your friends.’ Still, some data appear to indicate that French Tough-movement infinitives behave more like verbs than nouns. Chief among those is the fact that they can host pronominal clitics as (25) shows. (25) Cette bague serait difficile à lui voler sans qu’elle s’en aperçoive. this ring would-be difficult to her-DAT to-steal without that-she of-it notice ‘This ring would be difficult to steal from her without her noticing.’ However, as noted in Haïk (1985), who attributes the observation to Luigi Rizzi, there exists an unexpected difference between subcategorized and non-subcategorized clitics: Only subcategorized clitics can attach to Tough-movement infinitives as the contrast between (25) and (26b) shows. (26) a. Il sera facile d’y obtenir ce manuscrit. it will-be easy of-there to-obtain this manuscript ‘It will be easy to get this manuscript there.’ b. *Ce manuscrit sera facile à y obtenir. this manuscript will-be easy to there to-obtain ‘This manuscript will be easy to get there.’ We add to Haïk’s paradigm the case of affected dative clitics, which cannot attach to Tough-movement infinitives either, as shown in (27). Such dative clitics are not subcategorized and offer a nice contrast with their subcategorized counterpart in (25). (27) a. Il serait facile de (leur) endommager ces toiles. it would-be easy of (them-DAT) to-damage these paintings ‘It would be easy to damage these paintings (on them).’ b. Ces toiles seraient faciles à (*leur) endommager. these paintings woul-be easy to (*them-DAT) damage ‘These paintings would be easy to damage (on them).’ A possible explanation for such contrasts is that the subcategorized dative clitic in (25) is an agreement marker affixed to the gerundive, verbal noun in the lexicon rather than an independent functional head that amalgamates with a verb in the course of the syntactic derivation. On a movement analysis of Clitic Climbing in restructuring contexts we then

12

expect clitics subcategorized by Tough-movement infinitives in say, Italian, to be unable to climb to the tensed verb. This is, in fact, what happens, as shown in (28). (28) *Il libro gli fu difficile da offrire. the book him-DAT was difficult of to-give ‘The book was difficult to give to him (as a present).’ (example attributed to Luigi Rizzi by Haïk (1985:408)) Also consistent with the view that French Tough-movement infinitives are gerundive, verbal nouns is the fact that, unlike verbs, they cannot co-occur with adverbs of quantification such as always, as illustrated by the contrast in (29). Further, as shown in (30b), French Tough-movement infinitives are incompatible with floated quantifiers that correspond to their raised object, again contrasting with their verbal counterparts, as illustrated by the passive in (30a).4 (29) a. Il est impossible de toujours gagner cette course. it is impossible of always to-win this race ‘It is impossible to always win this race.’ b. *Cette course est impossible à toujours gagner. this race is impossible to always to-win ‘This race is impossible to always win.’

4 A reviewer points out that while (29b) is ungrammatical, the acceptability of frequency adverbs like toujours ‘always’ markedly improves in sentences like (i). (i) Ce mouvement est difficile à toujours exécuter parfaitement. this move is difficult to always to-execute perfectly ‘This move is difficult to always execute flawlessly.’ It seems to us, however, that the frequency adverb toujours ‘always’ in (i), unlike the one in (29b), modifies not the infinitive but rather, the manner adverb parfaitement ‘perfectly’. Since manner adverbs are compatible with both French Tough-movement infinitives and event nominals, it comes as no surprise that the combination toujours parfaitement is also acceptable with nominals as illustrated in (ii): (ii) L’exécution de cette maneuvre toujours parfaitement témoigne the-execution of this maneuver always perfectly attests d’un entraînement intensif. of-a training intensive ‘The invariably perfect execution of this maneuver is a testament to intensive training.’

13

(30) a. Ces livres ont été tous mis dans un seul tiroir. these books have been all put in a single drawer ‘These books were all put into a single drawer.’ b. *?Ces livres seraient faciles à tous mettre dans un seul tiroir. these books would-be easy to all to-put in a single drawer ‘These books would be easy to all put into a single drawer.’ (example from Kayne (1975:48, n.55) French Tough-movement infinitives are, however, compatible with manner adverbs, as (31a) shows, but this cannot be used as an argument for their verbal status since event nominals can be modified by such adverbs as well, as shown in (31b). (31) a. Cette maneuvre est impossible à exécuter lentement. this maneuver is impossible to to-execute slowly ‘This maneuver is impossible to execute slowly.’ b. Le nettoyage du canon lentement est nécessaire. the scrubbing of-the barrel slowly is necessary (afin d’enlever les résidus de poudre.) in-order of-to-remove the residue of powder ‘Scrubbing the barrel slowly is necessary (to remove powder residue).’ A last important question, raised by a reviewer, is why French Tough-movement infinitives, which we have argued are gerundive nominals, appear to differ from deverbal NPs in allowing wh-extraction, as the paradigm in (32)-(33) illustrates. (32) a. Un tel livre est facile à offrir aux enfants. a such book is easy to to-give to-the children ‘Such a book is easy to give to children.’ b. A qui un tel livre est-il facile à offrir? to whom a such book is-it easy to to-give ‘To whom is such a book easy to give?’ (33) a. Il critique l’enseignement prématuré des mathématiques aux enfants. he criticizes the-teaching premature of-the mathematics to-the children ‘He is against the premature teaching of mathematics to children.’ b. *A qui critique-t-il l’enseignement prématuré des mathématiques? to whom criticizes-he the-teaching premature of-the mathematics However, it has been known for some time that wh-extraction out of nominal phrases is, in fact, possible if they are indefinite, as (34) illustrates. (34) De quelles matières dispensera-t-on un enseignement rudimentaire? of which fields will-dispense-one a teaching rudimentary ‘Which fields will they teach a basic knowledge of?’

14

In recent years, contrasts such as (33b) versus (34) have been taken to mean that definite DPs, but not indefinite ones, are phases (see e.g., Radford 2004), hence wh-extraction out of definite DPs is ruled out as a violation of the Phase Impenetrability Condition. Given this, we would like to suggest that the licitness of wh-extraction in (32b) should be taken to indicate that French Tough-movement infinitives are indefinite nominal gerunds. In the next section, we will argue that FTM infinitivals are in many ways similar to English Retroactive Gerunds. One notable difference between the two is that unlike FTM infinitivals, English Retroactive Gerunds may be introduced by an overt determiner (cf. ft. 6 below). While we will have nothing to say about why this difference should exist, it is interesting to note that the determiner of English Retroactive Gerunds is always an indefinite one. 5. FTM infinitivals and English retroactive gerunds Huddleston (1971), Hantson (1984) and Clark (1985) discuss a type of English gerunds known as Retroactive Gerunds whose properties are strikingly similar to those displayed by FTMs. Retroactive Gerunds occur in sentences like (35). (35) a. Your leg needs amputating. b. Your book deserves translating into several languages. As noted by Hantson (1984), Retroactive Gerunds are often replaceable by a passive infinitive, as (36) illustrates.5 We thus have a first intriguing parallel between Retroactive Gerunds and Romance Tough-movement since, as shown in (21), the infinitive in Spanish Tough-movement constructions can also sometimes be replaced by a passive form. (36) a. Your leg needs to be amputated. b. Your book deserves to be translated into several languages. Just like FTMs, Retroactive Gerunds are selected by a small class of predicates. Predicates that can appear in FTMs include facile ‘easy’, difficile ‘difficult’, emmerdant ‘tough’, impossible ‘impossible’ but not, for example, possible ‘possible’, as shown in (37). Predicates that select Retroactive Gerunds, called “predicates of requirement” by Clark (1985), include need, want, deserve, merit, require, bear but not, for example, escape, as shown in (38).6

5 In some dialects, one finds past participles in the place of Retroactive Gerunds, mostly under the predicate need (i). The question of whether or not these are elliptical passives is beyond the scope of this paper. (i) a. Your hair needs cut. b. This bridge needs worked on. c. *Your book deserves translated into several languages. 6 Retroactive Gerunds can also appear with an overt indefinite determiner (i), in which case their status as nominal gerunds is obvious. Such Retroactive Gerunds have been called “retroactive nominals” by Clark (1985). They are selected by exactly the same class of predicates as determinerless Retroactive Gerunds and display all of the same properties discussed in the text.

15

(37) a. Il est maintenant possible de traiter ces cancers. it is now possible of to-treat these cancers ‘It is now possible to treat such cancers.’ b. *Ces cancers sont maintenant possibles à traiter. these cancers are now possible to to-treat (38) *The prisoner escaped torturing. The similarities do not end there, however. Just like FTMs, Retroactive Gerunds do not license unbounded dependencies as shown in (39a), nor do they license parasitic gaps, as shown in (39b), which suggests that they are not a case of wh-movement either. In this respect, Retroactive Gerunds contrast sharply with English-type Tough-movement induced by worth, as illustrated in (40). (39) a. *His leg needs persuading that surgeon to amputate. b. *This book does not need praising ___ without reading PG. (40) a. His leg is worth persuading that surgeon to amputate. b. This book is not worth praising ___ without reading PG. Further, Retroactive Gerunds and French Tough-movement infinitivals present striking similarities when it comes to the syntactic realization of the source of the event they denote as a by-phrase. This can be seen by comparing the examples in (41) with those in (23). (41) a. That point needs stressing by any linguist. b. This problem bears thinking about by everybody in the class. (examples taken from Hantson (1985)) In fact, Retroactive Gerunds, just like French Tough-movement infinitivals, are, in many ways, parallel to passives. For example, neither Retroactive Gerunds nor French Tough-movement infinitivals can occur with non-passivizable verbs, as the paradigm in (42) illustrates, and Retroactive Gerunds that involve “preposition stranding” are grammatical only with those prepositions that license passive movement, as shown by the contrast between (43) and (44).7 8

(i) a. John could use a good talking to (by a competent psychiatrist). b. These ideas merit some working on. c. This problem bears a good deal of thinking about. 7 Hantson (1984:103, n.3) explicitly objects to Huddleston’s (1971:153-154) analysis of Retroactive Gerunds as involving raising on grounds that they are impossible with idiom chunks. He gives the following examples: (i) *Headway needs making. (ii) *? Tabs need keeping on his operation.

16

(42) a. *The required credentials were had by John. b. *These credentials need having (by all applicants). c. Il est difficile d’avoir les qualifications requises. it is difficult of-to-have the qualifications requisite ‘It is difficult to have the requisite qualifications.’ d. *Les qualifications requises sont difficiles à avoir. the qualifications requisite are difficult to to-have (43) a. This plan deserves thinking about. b. This plan was carefully thought about (by the committee). c. These black marks on the wall need painting over. d. These black marks were painted over. (44) a. *Washington D.C. was gone to (by most of the students). b. *Washington D.C. requires going to (to obtain this type of clearance). c. *This soaking rain was worked after (by all the farmers). d. *This soaking rain definitely needs working after. In view of these facts, we would like to propose that FTMs and Retroactive Gerunds involve the lexical selection of gerundive, verbal nouns by a restricted class of predicates. These have the argument structure of nominals and cannot be associated with a Case feature. As a result, their object must be Case-checked by the closest available Case bearing element, which can be tensed T, as in (45), a higher V associated with objective Case, as in (46), or even French prepositional conjunctions such as avec ‘with’, as in (47). (45) a. The roast needs carving. b. Ce gigot d’agneau est difficile à découper. this leg of-lamb is difficult to to-carve ‘This leg of lamb is difficult to carve.’

It is not clear to us, however, that all idioms are thus excluded from the construction. As is often the case, “breaking up” idioms via raising leads to subtle and variable (along dialectal/idiolectal lines) judgments. Nevertheless, it seems to us that the example in (iii) is much more acceptable, thereby casting doubt on Hantson’s argument as a whole. (iii) ?They didn’t seem at all concerned that justice needed rendering in this case. 8 In contrast to our informants, one reviewer sharply disagrees with the judgment attributed to (42d), which s/he finds perfect. We have no explanation for this discrepancy. All we can suggest at this time is that perhaps this reviewer’s dialect allows for avoir ‘have’ to take on the meaning of obtenir ‘obtain’ in this context.

17

(46) a. She found the proposal to merit thinking about. b. Elle a trouvé Jérôme difficile à convaincre. she has found Jérôme difficult to to-convince ‘She found Jerome difficult to convince.’ (47) Avec mes amis si difficiles à contenter, ma vie n’est pas drôle. with my friends so difficult to to-please my life is not fun ‘Given that my friends are so difficult to please, my life is no fun.’ 6. Conclusion In this paper, we have argued for a new analysis of French Tough-movement infinitives as indefinite gerundive nominals. The empirical evidence presented supports the idea that, unlike other infinitivals, French Tough-movement infinitives are not clausal infinitivals dominated by a TP node but, rather, gerundive event nominals without a projected external argument that are possibly dominated by a non-overt indefinite determiner. In this respect, French Tough-movement infinitives are more like Retroactive English gerunds. Thus, what appears to be an infinitival morpheme (e.g., -er, ir, etc.) in French Tough-movement infinitives is, in fact, a nominalizing suffix.

The Pennsylvania State University References

Canac Marquis, Réjean (1996). The distribution of à and de in tough constructions in French.

In Grammatical theory and Romance languages, Karen Zagona (ed,), 35-46. Amsterdam: John Benjamins.

Chomsky, Noam (1982). Some concepts and consequences of the theory of government and binding. Cambridge, MA.: MIT Press.

Clark, Robin (1985). Boundaries and the treatment of control. Ph.D. Dissertation, University of California, Los Angeles. Déprez, Viviane. 1990. On the typology of syntactic positions and the nature of chains. Ph.D.

Dissertation, Massachusetts Institute of Technology, Cambridge. Folli, Raffaella and Heidi Harley (2007). Causation, obligation, and argument structure: On

the nature of little v. Linguistic Inquiry 38:197-238. Grimshaw, Jane (1990). Argument structure. Cambridge, MA.: MIT Press. Guasti, Maria Teresa (1990). The ‘faire-par’ construction in Romance and in Germanic. In

Proceedings of the Ninth West Coast Conference on Formal Linguistics, Aaron Halpern (ed.), 205-218. Stanford: CSLI Publications.

Haïk, Isabelle (1985). The syntax of operators. Ph.D. Dissertation, Massachusetts Institute of Technology, Cambridge. Hantson, André (1984). Towards an analysis of retroactive gerunds. In Sentential

complementation, Wim de Geest and Yvan Putseys (eds.), 95-103. Dordrecht: Foris. Huddleston, Rodney (1971). The sentence in written English. Cambridge: Cambridge

University Press.

18

Kayne, Richard (1975). French syntax: The transformational cycle. Cambridge, MA.: MIT Press.

Legendre, Géraldine (1986). Object raising in French: A unified account. Natural Language and Linguistic Theory 4:137-183.

Montalbetti, Mario and Mamoru Saito (1983). On certain (tough) differences between Spanish and English. In Proceedings of the Thirteenth Annual Meeting of the North East Linguistics Conference, Peter Sells and Charles Jones (eds.), 191-198. Amherst, MA.: GLSA publications.

Radford, Andrew (2004). Minimalist syntax: Exploring the structure of English. Cambridge: Cambridge University Press.

Travis, Lisa (1992). Inner aspect and the structure of VP. Cahiers Linguistiques de l’UQAM 1:130-146.

Williams, Edwin (1985). PRO and subject of NP. Natural Language and Linguistic Theory 3:297-315.