The Position of Clitics in Phrases with an Infinite Verb Form in Romance Languages

It is generally held that the original Indo-European word-order is SOV1, but this changed over time, and SVO is now a common feature of Indo-European languages. J.H. Greenberg argued that ”if in a language the verb follows both the nominal subject and the nominal object as the dominant order, the language almost always has a case system.”2 W. P. Lehman and others have long argued that prepositions have been introduced later to a VO + postposition system, thus undermining the case system by making the SVO order necessary to differentiate nominal subject from nominal object.3 However, some linguists have suggested that in Proto-Indo-European there also existed prepositions and therefore the SVO possibility as well.4 J. A. Hawkins has also stressed ”that the existence of VO & + Case languages, such as Lithuanian, means that the loss of the case system is not a necessary cause of the OV to VO shift, since OV can shift to VO whether or not the case system is eroded, but it is certainly a sufficient cause.”5 93

It is generally held that the original Indo-European word-order is SOV 1 , but this changed over time, and SVO is now a common feature of Indo-European languages.J.H. Greenberg argued that "if in a language the verb follows both the nominal subject and the nominal object as the dominant order, the language almost always has a case system." 2 W. P. Lehman and others have long argued that prepositions have been introduced later to a VO + postposition system, thus undermining the case system by making the SVO order necessary to differentiate nominal subject from nominal object. 3However, some linguists have suggested that in Proto-Indo-European there also existed prepositions and therefore the SVO possibility as well. 4J. A. Hawkins has also stressed "that the existence of VO & + Case languages, such as Lithuanian, means that the loss of the case system is not a necessary cause of the OV to VO shift, since OV can shift to VO whether or not the case system is eroded, but it is certainly a sufficient cause." 5 Modern Romance languages all have the SVO order but for clauses where clitics substitute nouns.
Especially in Europe the SVO order is sometimes considered to be "the logical order". 6Latin being a case system language was mostly a SOV language, although word-order was fairly free due to the existing case system (1a and 1b). 7In example (1a) we find both a noun and a clitic in the position of a direct object.The mere existence of the SVO possibility in Latin, even if it was the less used one, explains why the shift was so easily made and perhaps also why there is still a remnant of the old SOV system in Romance languages.Given that both word-orders coexisted the system change represented a change of preference in the predominant word-order rather than a total system shift.
Where there is a full noun object the normal modern Romance wordorder is SVO (2a and 2b).Topicalization changes this, although this anaphoric process lifts out one element of the standard composition of the phrase, and this has to be substituted by a clitic (2c and 2d).In this article I use the denomination clitic for a complementary pronoun without any inherent accent, although I will discuss a few examples where the clitics in the given sense have been substituted by stressed forms like in (3a).9Reflexives are, of course, also clitics and are thereby included in this discussion.However, I will not discuss clitic doubling (3b) or configurations with non-clitic reflexives where no clitic doubling is required (3c). 10Afirmative imperative forms are always constructed with enclitic pronouns -clitics or not (3a and 4a).In some vernaculars the enclisis is still found in spoken language such as in the Carribean, Chile and some other parts of Latin America (4b), where the plural ending is repeated after the clitic or because of a metathesis at the end of the finite verb after the enclitic pronoun. 11a)  It would seem reasonable to believe that Galician and Portuguese would admit the enclisis after a past participle, since the enclisis is the rule in afirmative clauses in these languages, but this is not at all the case.The enclisis is totally excluded in combination with a past parti-ciple (5a).13This is also the case in clauses where the present participle is allocated to the left (5b).When an object noun phrase is substituted by a clitic, the Romance standard order is SCV, where C stands for Clitic (6a and 6b).This order is equivalent to the original SOV.In medieval Romance texts both SCV an SVC are found (6c-f).The enclisis was the standard position in medieval Spanish, but uncertainty already appeared in the 16th century as can be seen in Cervantes' certificate of baptism (6g Another factor that also ought to be taken into consideration is that it seems just to believe that structural basics resist changes longer as E. Sapir once pointed out: "Languages are in constant process of change, but it is only reasonable to suppose that they tend to preserve longest what is most fundamental in their structure." 18 Romance languages clitic climbing is an old pattern that is spreading.This phenomenon has in some countries been seen as bad language and therefore ousted. "One assumes that there was considerable cohesion between the sequential verbs, so that they filled a single verb slot.The climbing pattern seems sporadically to have spread by analogy to other finite verb + infinitive sequences, even where the infinitive is introduced by a preposition (or complementizer).However, in the modern period there has been some reaction against climbing, with a tendency prompted by logical and semantic considerations, to cliticize the object pronouns to the verb to which it most closely relates.This process is most advanced in French, where climbing with modals is no longer permitted in the standard, following the intervention of the language arbiters." 19wever, Romansh is the Romance language that stands out, possibly due to German influence (7a) 20 , since Germanic word-order is SVC as can be seen in (7b and 7c).The Sursilvan does not even have clitics.
There are only full objective forms to be used.Engadine dialects hold to the Romance pattern as can be seen in the Vallader (Lower Engadine) example (7d).The central dialect, Surmiran, mostly follows the Sursilvan pattern (7e), although the clitic possibility still exists but is now restricted to literary usage (7f). 21Portuguese does at a first glance resemble Romansh since the regular word-order is SVC (8a and 8b).The difference is that in Portuguese the pronouns are clitics and they are clearly enclitically linked to the finite verb (8a and 8b), even prosodically and not just orthographically with a hyphen.This word-order is also found in Galician and Asturian (8c)./23 These contrasting word-orders can be seen in ( 8e).Since this enclitic pronoun can never be stressed, it is a completely different case from the Romansh SVC order.It would seem reasonable to regard this SV+C in order to point out the close relation between the finite verb and the clitic.This might be seen as a variety of the SCV order in the sense that the clitic is linked to the finite verb although by enclisis.
The +C is a way of denoting the enclisis.As the proclitic word-order is compulsory in many instances of affirmative main clauses European Portuguese does not provide a real counterexample to the SCV order being the general Romance word-order.In Brazilian Portuguese clitics are always proclitic (8f), and, as we have just seen, in certain clauses proclisis is the rule also in European Portuguese. 24In the Sursilvan (Romansh) case as in Germanic languages the postverbal non-clitic pronoun can be both stressed and separated from the finite verb by a negation or adverbial complement (8g).As we have already seen this is not the case in Engadine dialects (8h).Nevertheless, in colloquial Brazilian Portuguese there is a clear preference for full forms, and clitics are mostly avoided. 25The standard word-order in Brazilian Portuguese is still the proclisis (8f).The SCV word-order is also valid for compound tenses (9a and 9b) in all the Romance languages with the exception of some Romansh dialects that go an entirely different way, as we have already seen.Another minor exception is the third person singular feminine in Romanian, which is always positioned after the participle (9c).However, loro (as indirect object) in Italian is unique as it can be interposed between auxiliary and participle (9d), although the postposition is much more frequent (9e).On the other hand this pronoun is not a clitic. 26Affirmative imperatives always use the VO or VC word-order in the Romance languages (4a, 10a and 10b), and for obvious syntactical reasons the subjects are excluded from the sentences.
(10a) ¡Ábrela!(S) Clitic doubling is frequent in colloquial Romance languages, and has a topicalizational function.As has been shown earlier this anaphoric process lifts out one element of the standard composition of the phrase, and that element must be substituted by a clitic (11a and 11b).This will not be dealt with here, but for two interesting examples.In Romanian dialects (Transylvanian and Moldovan) the clitic can take both the proclitic and the enclitic position (12a).This can also be the case in Friulian (12b). 27In non-standard Transylvanian Romanian the auxiliary can be postpositioned and the clitic inserted (12c).This forms a foot, a phonetic group, where the second syllable of the participle carries the stress. 28A similar word-order can also be found in Sardinian where the finite verb (podere) is postpositioned (12d).29  Similar to the imperative is the Italian deictic adverb ecco which also postpositions the clitic (13a) as does the Portuguese eis/ei (13b).This, as well as the Portuguese tmesis, i.e. the interpolation of the clitics in the future and conditional (13c) as for example Spaniards used to do in the Middle ages (13d), is left out in the following discussion.It should perhaps be stressed that my intention is not to describe all possibilities and regional varieties in all Romance languages but to give a good idea of the complexity as well as general tendencies.The negation is put before the clitic(s) in modern Romance languages SnegCV, although this was not always the case in medieval times (14a).Modern Portuguese also accepts clitics positioned before the negation in interrogative sentences and certain subordinated clauses (14b).In Italian the proclisis used to be considered elegant in connection with a negation and an infinitive (15a). 31Some Romance languages like French and Occitan often add an extra complement (pas or ges etc.) to the negation and this comes directly after the finite verb (15b and 15c).Some Romansh dialects always stick to the postposition, and like English the clitic is positioned after the participle (15d and 15e).The question is now what happens in more complex sentences where there are at least two verbs of which one is finite and the other one is infinite.Since the verbal syntagm (VP) in these sentences consists of two parts -V f (=finite verb) and V i (=infinite verb) -the clitics could be attached to any of these two acting verbs had it not been for V f often being an auxiliary verb that does not present any action of its own but merely modifies the message presented in the phrase.It would therefore seem only logical if the clitic kept to the V i .This is by no means the case.On the contrary the clitic attaches more often to V f , as has been suggested earlier.
The notion auxiliary verb is in itself an often discussed and not very well defined category.As H. Kronning points out two main criteria must be fulfilled: it must take an impersonal mode like infinitive, parti-ciple and gerund, and it must have a grammatical signification, i.e. temporal ('have'), aspectual ('go to') or modal ('can').32That these verbs are truly part of the same VP, is shown by the fact that referentially they constitute one unit and, consequently, an answer can refer to the whole VP by using only the auxiliary without the usage of any clitic (16a-c).If the finite verb is not an auxiliary the presence of a clitic is necessary in the answer, provided this is not just a simple 'yes' or 'no' (16d and 16e).From a strictly formal point of view it is the conjugated finite verb that is the nucleus of the VP, which implies that this is the central part of the message of the VP, and it should be a very natural thing that this fundamental element of the syntagm should be the one to gather other parts of the syntagm like clitics.However, from a semantic point of view the central message lies with the infinitive, i.e. infinite verb.The finite modal (and to a certain extent the aspectual) auxiliary only modifies the emission of a message on a pragmatic level.It does not change the essence of the message but the way this is presented.
In phrases with an auxiliary and an infinite verb form there are four possible positions for the clitic: SCV f V i , SV f +C V i , SV f CV i and SV f V i C (17a-d).As we have seen in ( 12c) and (12d) there also exists the marginal possibility of a V i +C(+)V f order, but this is more surprising from a verb order point of view.The position of the clitic is actually the same as in SV f V i C.
The space between two of the letters in the latter combinations is there to point out which verb is attracting the clitic, when this is inserted between the two verbs.The + is to describe the position of an enclitic pronoun like in (17b).In the Romanian case the infinite constructions are rare 33 , and mostly substituted by an infinitive like in many other languages that belong to the Balkan "Sprachbund" (17e).Certain verbs like putea, 3 stie, etc. 34 , may still use the infinitive that exists, although it is seldom used (16c), since the subordination is predominant.In French the word-order is now fixed SV f CV i, (18a), although the proclitic position was preferred in old French (18b) and existed as a possibility in the 17th century (18c). 35That both existed at the same time is clear from the example where both positions are included in the very same proposition.Although Sardinian can insert the clitic between the verbs (12d), the proclitic position also exists (18d).In other Romance languages the SCV f V i (and SV f +C V i in the Portuguese, Galician and Asturian cases) and SV f V i C coexist.Albeit the latter has been promoted, the modern tendency is that the clitic precedes the finite verb. 37The SOV order is thus practically the only basic order in Romance languages when the object is a clitic, with the already mentioned Portuguese exception SV f +C.Sursilvan and colloquial Brazilian Portuguese are disregarded since they do not use clitics.
It seems that SOV is becoming the predominant word-order in phrases with two verbs.Thus the general SVO system is out-ruled by what is the standard word-order in constructions with clitics, SOV.In phrases where there are two verb forms there is a vacillation between the two systems.
As has been shown French was to develop a SV f CV i system, but there is an exception where the SCV f V i is still applied (20f and 21a).Infinitive is very rare in Romanian and only appears after a very limited number of verbs, which means that the SCV stays valid, although it is extended to SV f Q, where Q = conjunction (sa*/ca*) + CV f .(19a;cf.19b with SV f V i C and 19c SCV f V i ).In the Standard Romanian construction there are two finite verbs which differentiate it from all other Romance languages in sentences where the two verbs take the same subject.The competition between the structurally fundamental SVO order and the SCV order that could be seen as a remnant from an older Latin system -SOV, has its consequences in sentences with two verbs, since this opens the three possibilities SCV f V i (20a and 20b), SV f CV i (20c) and SV f V i C (20d and 20e).Only French can take the second order, and this is now the standard French order.Nevertheless, the first order is compulsory in combinations with certain verbs (20f-h). 38Posner maintains that modal verbs have a preference for climbing, and this seems to be the case, although it is still only a preference. 39 will argue that since the SCV order has come to be the rule in most Romance languages this tends to prevail over the basic SVO order in SV f V i phrases.That this is the case is motivated by the fact that in constructions where there is an option: both SCV f V i 40 and SV f V i C are possible, but when only one word-order is possible it is almost always SV f V i C that is excluded.This means that the clitic microsystem, SCV surpasses the general SVO order.In the spoken languages this tendency seems to be growing 41 .French is the obvious exception, but historically, SCV f V i is also valid for French.In modern French SV f CV i is the generalised standard.Yet the SCV f V i is still required in some constructions, and these coincide with the cases in other Romance languages where SCV f V i is obligatory.
As we have seen the auxiliary is defined as a verb that must be constructed with an impersonal mode (infinitive, past participle or present participle), and it must have a grammatical significance. 42When this is not the case the two verbs do not constitute one single verbal action but two.However, in both cases we are dealing with what is traditionally described as accusative with infinitive (Accusativus cum infinitivo).The original meaning is final although this has been weakened."In dem Satz doceo te sapere kann der Inf.Sowohl von seiner urspr.finalen Bedeutung her verstanden werden ( " ich unterrichte dich zum Weisesein, damit du weise wirst") auch als Stellvertretener eines Sachobjekts ohne finalen Nebensinn neben dem persöhnlichen Objekt te wie in doceo te sapientiam ( " ich lehre dich Weisheit").Keine finale Auffassung des Infinitivs ist dagegen mehr möglich bei einem untergeordneten Verbum der Wahrnehmung wie video te venire, wo der Inf.zwar auch ein Sachobjekt vertritt, [...] Ganz eng schliesslich wird die Einheit, zu der Akk.und Inf.sich verbinden, wenn keiner von beiden mehr für sich als Objekt von dem überordneten Verbum abhängig gedacht werden kann [...] Je mehr nun Inf.und Akk.sich aus der Beziehung als selbständiges Objekt zum übergeordneten Verbum lösen und zusammenwachsen, desto leichter können sie gleichsam wie Prädikat und Subjekt eines entsprechenden Nebensatz escheinen:" 43 The origin of this construction in Latin referred to full verbs and not to auxiliaries.The same goes for causative Verbs in Latin 44 .N.B. the (S) 105 40 The SV f +C V i order is here regarded as a variety of SCV f V i .as it does not change the argument of this article. 41M. Casado, El castellano actual: usos y normas, 2nd ed. Pamplona, 1991, p.68. 42 H.Kronning,op. cit.,p. 59. 43 H. Rubenhauer and J.B. Hofmann,op. cit.,p. 192. 44 Ibid.
V f CV i order in the examples in the quotation corresponds perfectly to modern French.
Although there is a clause union between the two verbs that makes them a syntactic unit, there is a constructional difference between two full verbs (I) on one hand and an auxiliary + a full verb (II) on the other.Syntagmatically the predicate consists of two verbal actions (VP) that are interdependent in the first case, while in the second case the auxiliary does not express a verbal action of its own and depends on the full verb.This difference could be described in the following way: The concept of modal auxiliaries is fairly controversial and it could be argued that they should be included in category I.However, in combination with clitics they behave in a way that suggests their being considered parts of the same category as other auxiliaries, i.e. category II.
Causative or operator verbs, do and let, are more restricted in their usage then most verbs.In French the norm is SCV f V i (21a), while the standard order is SV f CV i as we have seen (21b). 45Nevertheless, the most frequently used construction is splitting the two clitics SCV f V i C (24c).Other romance languages seem to exclude SV f V i C or at least avoid it in conection with causative verbs (21c-21e). 46The V i enclisis is not accepted by native speakers when I have presented them all possibililities (21f and 21g).There is mostly a choice between an infinitive and a subordinate clause after a causative verb.However, this choice is semantic and when the meaning is 'make'/'do' the infinitive is often compulsory (22a), but when the meaning is 'see to'/'take care of' the subordinate clause is the first choice (22b), and ( 22c) is regarded incorrect. 47In these sentences there often exists a possibility of separating the two clitics.The general principle of always keeping the clitics together (23a and 23b) can thus be violated for semantic reasons (23c).Example (23c) shows that different subjects can make the clitics attach to both verbs.This is partly due to a close semantic relation between the clitic and the verb.In these sentences one clitic is object to V f and subject to V i and the other object to V i (24a -24e).In these sentences the V f is either a causative or a perception verb. 48This is not the case when V f is an auxiliary.If the two verbs take different subjects the V i has to be turned into a subordinate clause (24f and 24g).Sentences like (24h and 24i) are simply not used.To this category also belong verbs for 'know' and 'can' in their auxiliary meaning and function (26a-g).In these cases there is always a coreference between the subject (formally expressed or not) of the infinitive and one of the arguments, i.e. subject or object, of the finite matrixverb.These verbs are, as we have seen, among the few Romanian verbs that can actually be combined with infinitives (26d).In Italian sapere (and volere) mostly go with enclisis (cf.19b and 20d) while potere and dovere have clitics in the proclitic position (25c). 54n Portuguese there is a clear difference between European and Brazilian usage.In the latter case there is normally a SV f CV i order (26e), while the former takes either SV f +C V i or SV f V i C (26f).The proclisis is evidently favoured also in less well-known Romance languages like Dalmatian (26g).There are, of course, other kinds of full verbs that are attached to an infinitive with or without a preposition.The evidence points to a hierarchy of preference for climbing among verbs that take the infinitive, with modals and aspect auxiliaries high on the list, followed by conative (e.g.'try') (27a, 27b and 27h), motion (e.g.'come') (27c, 27f and 27g) and raising (e.g.'seem') verbs (27d). 56In French the gérondif is a separate case (27e).However, it is hard to find good examples of motion verbs in a concrete sense that take an infinitive, since they mostly take a gerund indicating a simultanous action (27f and 27g).It is also possible that the presence of a preposition and adverbs between the finite and the infinite verb forms promotes the enclisis if the clitic is semantically closely related to the verb (27h and 27i).
There also exists the possibility to expand the VP to two (or perhaps even more) infinite verb forms SV f V i V i .In this case the clitic could be linked to any of the three verb forms.However, it seems that the same tendencies, discussed earlier, apply to these clauses as well (30ac).Infinitives have come to be used a lot more in Romance languages than they were in Latin.As a consequence of this increased usage new syntactical structures have arisen.The problem that concerns us in this article is the location of clitics in sentences with two verbs of which one is finite and one if infinite.As we have seen there exist four possibilities: SCV f V i , SV f +C V i , SV f CV i and SV f V i C. The latter two correspond to the ruling word-order in Romance languages SVO, while the first reflects the SOV order (and so does the second).However, clitics preserve this latter word-order in all Romance languages with the exception of some Romansh dialects.Where there are two verbs, clitics can take either position.Nevertheless, the presumed free choice between these options turns out to be much more restricted than is generally held to exist between the proclisis and the enclisis.Causative and perception verbs all tend to favour the proclitic word-order, to the degree that this is not only preferred but in some languages the only possible one.In the spoken language this proclitic tendency is strong and spreading.The usage of certain verbs and syntactically interpolated elements like prepositions and adverbials must be considered in every single case, since these might have an influence on each occurrence, but this does not obscure the general view that the SCV order is favoured.We could thus conclude that although clitics are used both enclitically and proclitically in a fairly complex way, the SCV order is the dominant word-order in Romance languages also in sentences with two verbs of which one is finite and one infinite.
).Even today the enclisis is, of course, not entirely unusal in poetry and archaic language.