The Origin of the Portuguese Inflected Infinitive
Kenneth J. Wireback
Abstract: Two main theories have been proposed to account for the origin of the Portuguese inflected infinitive. One, the imperfect subjunctive theory, claims that the inflected infinitive developed from the Latin imperfect subjunctive in volitional clauses. The other, the creative theory, views the inflected infinitive as a strictly Luso-Romance phenomenon, without any direct Latin predecessor. In this study I adapt the imperfect subjunctive theory to include purpose clauses rather than volitional clauses to account for Old Portuguese data that show a high correlation between the purpose prepositions pera and por, and the inflected infinitive. Since the inflected infinitive is already established by the time of the earliest documents in Portuguese, there is no hard evidence that would permit a definite solution to the riddle of its development. Thus, this article is a reconstruction of changes based on attested syntactic patterns in Late Latin and Old Portuguese.
Key words: infinitive (personal), infinitive (inflected), Latin pluperfect subjunctive, Latin imperfect subjunctive, purpose clauses, Portuguese linguistics
One salient feature of Portuguese is the ability to inflect the infinitive in a variety of subordinate clauses. (110) First, the inflected infinitive may occur after impersonal expressions where the subject clause containing the personal infinitive has been postposed, e. g., É imprescindível partirmos já, 'it is absolutely necessary for us to leave now' (Thomas 1969: 189). Second, it may occur in declarative predicates: Garantiu os livros estarem no porão, 'he guaranteed the books would be in the hold (of the ship)' (Koike 1983: 93). Third, the inflected infinitive may be used following verbs of perception, e g., vi os rapazes fazerem a comida, 'I saw the young people make food' (Koike 1983: 93). Fourth, it may occur after verbs of emotion: Eu lamento os deputados tem trabalhado pouco, 'I regret that the deputies have hardly worked' (Raposo 1987: 87). Finally, a common use of the inflected infinitive is after a preposition: Ele abriu a cancela para os cavalos entrarem no curral, 'he opened the gate for the horses to enter the lot' (Thomas 1969:189). (111)
One can delineate two main theories on the origin and development of the Portuguese inflected infinitive. The first theory claims that the inflected infinitive developed from the Latin imperfect subjunctive in volitional clauses. The second theory, the creative approach, views the Portuguese personal infinitive as an extension of the uninflected infinitive in Old Portuguese, with no direct development from any aspect of Latin syntax (Osborne 1982: 243). In this article, I discount the creative theory and instead modify the imperfect subjunctive theory by locating the origin of the inflected infinitive in purpose clauses, rather than volitional contexts, in order to account for Old Portuguese data that show the predominance of postprepositional inflected infinitives with a purpose function.
This study is by nature hypothetical. As far as derivation from the Latin imperfect subjunctive is concerned, there is no concrete data that provide an intermediate link between the Latin imperfect subjunctive and the inflected infinitive. Similarly, by definition there can be no examples of any intermediate steps in the creative theory; the infinitive is either inflected or not. To overcome this, the creative theory must show what was inside the mind of speakers of Old Portuguese, which is ultimately impossible, in spite of certain claims to the contrary with regard to Portuguese speakers'  eagerness for verbal inflection (Moffatt 1967: 39-40). Because of these constraints, the exact origin of the inflected infinitive in Portuguese will probably always resist a definitive discovery (Osborne 1982: 245).
2. The Latin Imperfect and Pluperfect Subjunctives
Since the imperfect subjunctive theory derives the inflected infinitive from the Latin imperfect subjunctive, I will begin with an overview of the Latin subjunctive in the past tense. In Latin, one of two subjunctive verb forms could be used if the verb in the matrix clause was in the past tense, either the imperfect subjunctive for an action that was not completed with respect to the temporal reference point of the matrix clause, or the pluperfect subjunctive if the action was seen as completed in relation to the verb in the matrix clause (Bennett 1955: 171):
|Vidi quid faceres, 'I saw what you were doing' (imperfect subjunctive), vs vidi quid fecisses, 'I saw what you had done' (pluperfect subjunctive).|
The following two paradigms show the first conjugation (-are) forms of the imperfect and pluperfect subjunctives for amare, �to love�:
|amavissem||'I had loved'|
|amavisses||'You had loved'|
|amavisset||'s/he had loved'|
|amavissimus||'we had loved'|
|amavissent||'they had loved'|
According to Ernout and Thomas, the imperfect subjunctive is the older of the two forms, the pluperfect subjunctive still relatively rare during the Old Latin period (1953: 244). By imperial times, however, the pluperfect subjunctive is in the process of ousting the imperfect subjunctive as the primary past tense form of the subjunctive, due to certain non-past uses of the imperfect subjunctive. For example, in hortatory and deliberative subjunctive contexts, the imperfect subjunctive is used where one would expect a present subjunctive (Ernout and Thomas 1953: 244), e g., quid facerem? in place of quid faciam? �What am I to do?� Quid fecissem would then fill in the past tense slot vacated by the imperfect subjunctive and thus come to mean 'what was I to do', with the result that the imperfect subjunctive was associated with the present tense and the pluperfect subjunctive was the sole form for the past tense.
Other non-past uses of the imperfect subjunctive may be seen in conditional sentences. According to Bourciez (1946: 129) and Harris (1978: 236-37), the imperfect subjunctive had both past and non-past applications, depending on the impossibility or improbability of the hypothetical action. For example, si veniret, non me videret, 'if he did came, he wouldn't have seen me,' refers to a past time frame and states that it is improbable that he came, but there is still a possibility that he did. In contrast, si veniret, non me videret, 'if he came, he would not see me,' in effect states that either at the present time or in the future, it is impossible for him to come. The use of the imperfect subjunctive, therefore, is unstable in at least two ways. On the one hand, it can cross tense boundaries and function in both past and non-past contexts. On the other hand, it can be utilized in both improbable and impossible contexts (Harris 1978: 23637). This temporal instability of the imperfect subjunctive, plus the pluperfect subjunctive's rather salient phonetic characteristics with respect to the imperfect subjunctive, resulted in the emergence of the Pluperfect subjunctive as the primary past tense subjunctive form (Harris 1978: 163). By the third century A. D., the pluperfect subjunctive is the dominant form, and the imperfect subjunctive eventually disappears, ceding its environments to the pluperfect subjunctive, which becomes the imperfect subjunctive of the Romance languages  (Grandgent 1907: 53; Väänänen 1988: 233). (112)
3. The Imperfect Subjunctive Theory and Volitional Clauses
According to one theory, the imperfect subjunctive survived in Luso-Romance to become the Portuguese inflected infinitive. Essentially, this theory revolves around three main issues: the survival of the Latin imperfect subjunctive in Portuguese territory, phonological developments that link the imperfect subjunctive with the inflected infinitive, and the change from a tensed to an untensed form. For a full treatment of the theory that derives the inflected infinitive from the imperfect subjunctive, the reader is urged to consult Williams (1962: 181-84), and Maurer (1968: 7-19). Here I will just sketch the main points.
The primary task for a theory that derives the inflected infinitive from the Latin imperfect subjunctive is to document the survival of the imperfect subjunctive in Luso-Romance long enough for it to develop into the inflected infinitive. According to Rodrigues, 11th century Medieval Latin examples from Portuguese territory prove the survival of the imperfect subjunctive, as shown in the following example (1932: 3-4):
|-2||Et intrarunt in placito testimoniale pro in tertio die darent testes sicunt et fecerunt.|
|'And they began the hearing in order to, on the third day, provide witnesses, and they did this'.|
Rodrigues' thesis is that pro darent, 'in order to provide,' is the equivalent of Classical Latin ut darent, 'in order that they provide' (imp. subj.) (1932: 4). Although its use after pro suggests that syntactically darent is an infinitive with person/number endings, morphologically darent is definitely an imperfect subjunctive form. In short, for Rodrigues the occurrence of Latin imperfect subjunctive verb forms in Luso-Romance territory proves its survival into the initial stages of Old Portuguese.
Phonological developments also support the theory that derives the inflected infinitive from the imperfect subjunctive. For the first, second, and third-person singular forms, as well as the third-person plural forms, the phonological changes would be completely regular (Osborne 1982: 245-46):
|-3||amarem||>||amar,'I love (d)'|
|amares||>||amares, 'you love (d)'|
|amaret||>||amar,'s/he love (d)'|
|amarent||>||amarem, ''they love (d)'|
Loss of final /m/ and /t/ is well documented in Ibero-Romance, as is the apocope of final /e/. In contrast, the first and second person plural forms are not so straightforward. Normal phonological development would have produced *amaremos and *amaredes from amaremus, amaretis, respectively.
As Rodrigues indicates, however, changes in the primary word stress of these forms precluded their natural development (1932: 3), with the result that the stress in amaremus, amaretis was retracted one syllable to [amáremus] and [amáretis]. In this fashion, the unstressed posttonic /e/ was deleted, producing the personal infinitive forms amarmos and amardes, respectively.
The developments would be the same for the other two verb paradigms, as shown in (4) with tenere > ter, 'to have,' and venire > vir, 'to come':
|tenerem > ter||venirem >vir|
|teneres > teres||venires > vires|
|teneret > ter||veniret > vir|
|teneremus > termos||veniremus > virmos|
|teneretis >terdes||veniretis > virdes|
|tenerent > terem||venirent > virem|
The third cornerstone of the imperfect subjunctive hypothesis is the change from a tensed subjunctive form in Latin to an untensed, inflected infinitive in Luso-Romance. Through regular phonological change, the Latin imperfect subjunctive began to look like an infinitive with person/number endings, so that a complete coalescence of the two forms is possible provided that they share the same syntactic context. According to Gamillscheg, such a coalescence occurred in volitional clauses after the subordinating conjunction ut. The conjunction ut was often dropped in Latin, so  that placuit ut traderet, �it was agreed that he would bequeath�, could become placuit traderet. With the loss of the weak final /-t/, the infinitive tradere and the imperfect subjunctive traderet would then be identical, and other forms like traderes, traderemus, and traderent would appear to be nothing more than infinitives with person/number endings (Gamillscheg 1970: 266-67).
Since the Latin imperfect subjunctive does not survive as a tensed subjunctive form in Modern Portuguese, there can be no direct verbal development in this respect from Latin to Portuguese. Instead, the imperfect subjunctive theory posits a transfer of Latin imperfect subjunctive person/number morphology over to the uninflected infinitive once the imperfect subjunctive began to look like an infinitive with inflections. Therefore, the origin of the inflected infinitive is the union of the imperfect subjunctive's morphology with the uninflected infinitive's syntactic contexts.
The primary inconsistency of the imperfect subjunctive theory based on volitional clauses is the lack of correspondence between contexts employing the imperfect subjunctive in Latin and contexts employing the personal infinitive in Portuguese. Since the contexts for the coalescence of the uninflected infinitive and the imperfect subjunctive forms are volitional in nature, the inflected infinitive should appear in these same contexts in Old Portuguese. This, however, is not the case. In volitional constructions where ut ellipsis was common in Latin, Portuguese may not use a personal infinitive construction (Maurer 1968: 53):
|-5||rogavi (ut) scriberes|
|�I asked that you write�.|
The imperfect subjunctive theory cannot satisfactorily account for the development from the Latin subjunctive form to the inflected infinitive unless it locates the morphological transfer in a syntactic context that permits the inflected infinitive in Old Portuguese.
4. The Creative Approach
In view of this inconsistency in the imperfect subjunctive theory, the creative approach instead posits a strictly Luso-Romance development due to the use of lexical subjects with the infinitive and formal similarity with the future subjunctive as the origin of the Portuguese inflected infinitive. First, as Maurer observes, if we define a personal infinitive as an infinitive with a lexical subject, then other Romance languages, not just Portuguese, have personal infinitives, e g., Spanish mas por mandarlo vos, padre, lo haré, �but because you ordered it, father, I will do it� (1968: 69-74). The difference is that only Portuguese has an inflected infinitive, i. e., an infinitive with person/number endings. With the appearance of a lexical subject alongside the infinitive, there was a tendency to integrate the infinitive into the pattern of the finite verb forms, which are inflected for person and number (Maurer 1951: 37-39). This extra step occurs in Portuguese and not in the other Romance languages because only in Portuguese is the future subjunctive formally identical to the inflected infinitive in regular verbs. This resemblance, plus the existence of the lexical subject, extends the person/number endings of the future subjunctive to the infinitive. Then by analogy the endings are extended to the infinitives that would have irregular forms in the future subjunctive (Maurer 1951: 39; Martin 1960: 342).
The creative theory also suffers from certain inconsistencies. Three criteria that any theory on the origin of the inflected infinitive must obey are: 1) that there be another verb form with which the uninflected infinitive can be related analogically; 2) that there be a similar syntactic context where both the impersonal infinitive and the other verb form occur in order to allow the morphological transfer; 3) that this syntactic context also exist in Old Portuguese. (113) The imperfect subjunctive theory cannot satisfy the third requirement, because the context of the morphological transfer from the imperfect subjunctive to the uninflected infinitive, volitional clauses, do not permit the  inflected infinitive in Old Portuguese. More importantly, the creative approach cannot satisfy the second requirement, because the future subjunctive and the inflected infinitive do not share the same syntactic environment.
If an extension of the future subjunctive's person/number morphology to the uninflected infinitive is postulated, there must be cases in which the future subjunctive and the uninflected infinitive form the syntactic equivalent of a minimal pair -differentiated only by verb morphology and perhaps a conjunction. There is, however, very little overlap in this regard. Since the future subjunctive occurs after adverbials like quando, 'when,' and logo que, 'as soon as', and after the conjunction se, 'if', whereas the inflected infinitive does not, one of the few shared contexts available is following depois, 'after.' If the future subjunctive extended its morphology to the uninflected infinitive, this would be the only context in which the extension could occur. Although it is possible, only one context seems an unlikely origin to such a widespread verb form.
5. Inflected Infinitive Usage in Old Portuguese
Since two salient aspects of the above theories are the importance of syntactic context and the importance of a lexical subject, the latter especially with regard to the creative theory, one solution to the origin of the inflected infinitive is to analyze inflected infinitive usage in Old Portuguese in order to discover the most prevalent patterns of usage, and then develop a theory based on this data. From Portuguese works dated circa 1400 or earlier, an analysis of the syntactic context of 153 sentences employing a personal or inflected infinitive showed the following results. (114)
|Table 1: Contexts introducing the inflected infinitive|
|(sem, até com, em)||10||
|(extraposed subject, 7 after dever, depois)||
The first pattern that is apparent in table 1 is the predominance of a preposition as the element that introduces a personal infinitive, as shown in the following examples from Carter (1967):
|-6||E Por tanto te escolho pera pregares omeu nome.|
|'And therefore I choose you to preach my name.' (96)|
|Duvydavam de perderem ho que tynhã...|
|'They doubted they would lose what they had...'(356)|
|Jazião hum sobre outro sem se poderem mais fazer|
|'They lay one on top of the other without being able to do any more.' (276)|
Without distinguishing between prepositions, the total percentage of occurrences with a preposition comes to 95.4%, so it is clear that use of the inflected infinitive with a preposition is the dominant construction. The second pattern that is apparent is the predominance of the prepositions pera and por, �in order to �.They introduced the personal infinitive more than twice as often as de, the second most common preposition to precede the personal infinitive. This preference for pera/por suggests that the origin of the inflected infinitive lies in a purpose construction, since pera and por essentially establish a purpose relationship between two clauses. (115)
Neither the imperfect subjunctive theory nor the creative theory, as they stand now, is capable of accounting for this propensity to occur with a preposition, especially pera/por. The contexts cited by Rodrigues (1932: 3-4) and Gamillscheg (1970: 26668), that originated the morphological extension from the Latin imperfect subjunctive to the uninflected infinitive, are volitional in nature, and the imperfect subjunctive theory has no way to link these volitional contexts with the prepositional contexts of Old Portuguese. Even if we were to allow certain prepositional examples to come under the  definition of volitional contexts, e g., e dise lhes senhores vos bem vedes como vosha mister deserdes boõs..., �and he told them, men, you clearly see how it is necessary for you to be good...� (Carter 1967: 134), the imperfect subjunctive theory cannot explain why these volitional uses are limited to prepositional contexts. (116)
Similarly, the creative approach cannot explain this preference for prepositional contexts. The future subjunctive does not occur after prepositions, so postprepositional usage could not have been transferred with the morphology from the future subjunctive. Furthermore, the one syntactic context in which this transfer could have occurred, after depois, appears not to have been very prevalent. Out of the 153 sentences analyzed in the present study, only one sentence shows the inflected infinitive after depois: Vyerã aly depoys muitos cavaleiros provare[m] pera jazere [m] hy de noute... �they came there after many knights had tried, in order to lie down there at night...� (Carter 1967: 371). If future subjunctive morphology were the originator of the inflected infinitive, then sentences with depois with the inflected infinitive should be much more common.
Another contextual factor employed by the creative theory is the occurrence of a lexical subject with the infinitive in Old Portuguese. In fact, Maurer claims that this is the primary reason for the inflection of the infinitive. According to him, once the infinitive became �personal�, i. e., took a lexical subject, it was only natural that it fall in line with other finite verbs that agree with their subject, thereby becoming an inflected infinitive (1951: 39-40). If this hypothesis is true, then an examination of inflected infinitive usage in Old Portuguese should show a high occurrence of overt lexical subjects. As shown in table 2, however, the opposite seems to have been the norm:
|Table 2: Inflected infinitive with lexical subjects|
|no lexical subject||138||90.2|
|with lexical subject||15||9.8|
It is clear from the data presented in table 2 that an overt subject in the same clause occurs less than 10 percent of the time, so it seems unlikely that the subject could have been present with enough frequency to cause the inflection of the infinitive.
The above analysis of inflected infinitive usage in Old Portuguese creates problems for both the imperfect subjunctive theory based on volitional clauses and the creative theory. The imperfect subjunctive theory does, however, provide a solid basis for a solution to the phenomenon, because it employs syntactic contexts in which the impersonal infinitive and the imperfect subjunctive may alternate, whereas the creative theory does not. Thus, I take the imperfect subjunctive to be the origin of the inflected infinitive. The problem is to link this point in the development to the usage detailed above in table 1. Specifically, why should the inflected infinitive occur so often with a preposition, especially with pera and por? To account for this postprepositional usage, I develop a solution within the general framework of the imperfect subjunctive theory, but based on purpose clauses rather than volitional constructions.
6. Imperfect Subjunctive Theory and Purpose Clauses
The data introduced in the previous section indicate that the dominant context for the inflected infinitive in Old Portuguese is after prepositions that establish a purpose relationship between independent and dependent clauses. In view of the fact that the imperfect subjunctive is the more plausible origin of the inflected infinitive, in morphological and syntactic terms, the solution to the origin of the inflected infinitive must find a common link between the Latin imperfect subjunctive, infinitive constructions, prepositional constructions, and purpose  constructions.
With respect to the Latin imperfect subjunctive, Ernout and Thomas state that it was the primary form of the subjunctive in purpose and result clauses when the matrix verb was in the past tense, since in a purpose or result construction the verb indicating the intended goal or result is necessarily subsequent to the action of the independent clause, e g., id tibi dedi ut memor esses, 'I gave it to you so that you would be mindful of it', where being mindful is subsequent to my giving (1953: 342-45, 414-17). It is possible then, that the Latin pluperfect subjunctive could more easily displace the imperfect subjunctive in other constructions where the imperfect subjunctive was not the favored verb form, while in purpose and result constructions the imperfect subjunctive may have survived longer.
Although the pluperfect subjunctive had become the primary form by the third century A. D., e g., timui ne inter nos bella fuissent orta, 'I was afraid lest the wars break out between us' (Bourciez 1946: 129), textual evidence indicates that the imperfect subjunctive continued to survive, at least in purpose and result clauses. For example, in the Peregrinatio ad Loca Sancta, the imperfect subjunctive is used in the following result clauses (Harrington 1962: 2-4):
|-7||Iter sic fuit ut per medium transpersaremus caput ipsius vallis et sic plecaremus nos admontem Dei.|
|�The journey was made in such a way that we crossed through the middle of the head of that valley and thus arrived at the mountain of God�.|
That the imperfect subjunctive should appear in a purpose or result clause is important, in view of the fact that the Peregrinatio is believed to represent:
|... the speech of a comparatively uneducated provincial towards the close of the fourth century, exhibit[ing] already in diction, signification of words, syntax, and style numerous features of the decadence of the language and its progress towards the Romance dialects which were its heirs. (Harrington 1962: 1).|
Instead of ut plus the imperfect subjunctive in purpose clauses, Latin also employed the infinitive, especially in Early and Late Latin: Eximus... ludos videre, �we are going... to see the games� (Harris 1978: 196); venerat aurum petere, �he had come to ask for his gold� (Ernout and Thomas 1953: 260-61); vado piscari, �I go a fishing� (Bourciez 1946: 110). As Harris observes, such a construction is to be expected given that the infinitive often occurred in accusative case as the complement of a verb (timeo abire, �I'm afraid to leave�), and that the accusative case originated as the marker of goal (i. e., the purpose for an action): Romam ire, �to go to Rome� (1978:40,19697).
At a certain stage in the history of Luso-Romance, therefore, there were at least two ways to indicate a purpose relationship, ut plus the subjunctive or an infinitive. Since the Latin conjunction ut was often deleted, e g., placuit ut vinderemus placuit vinderemus, �it was agreed that we sell� (Gamillscheg 1970: 267), the stage was set for the reanalysis of the imperfect subjunctive as an infinitival form and the extension of the imperfect subjunctive's morphology to uninflected purpose infinitives. With regular phonological development, the imperfect subjunctive began to look like an inflected infinitive, and with the elimination of ut could be reanalyzed as an infinitive with inflectional morphology. Furthermore, since uninflected infinitives were also employed in purpose contexts, the established purpose-infinitive pattern could make it even more likely that a reanalysis of the imperfect subjunctive form take place. By postulating the development in this way, we provide the needed link between infinitives, purpose clauses, and the Latin imperfect subjunctive. Although the present analysis is very similar to the imperfect subjunctive analysis based on volitional clauses, the advantage to postulating the morphological extension in a purpose context is that it accounts for the purpose function seen in Old Portuguese.
The subsequent use of these infinitive forms with purpose prepositions, which is well documented in Old Portuguese, is related to the extension of infinitive usage to prepositional contexts in Vulgar Latin and Early Romance. To get to the stage where  any type of infinitive may be used with a preposition, Harris proposes the following chronology of Romance development (1978: 196-97). First, by the Classical Latin period, the present active infinitive had already come to be used in nominal functions, e g., as subject with esse, �to be�, and as the object of a verb: Errare est humanum, �to err is human�; timeo abire, �I'm afraid to leave�. Since nominals also occur postprepositionally, a logical extension of this nominalization of infinitives is as the object of a preposition (Harris 1978: 197).
Once this preposition-plus-infinitive patterns was established, the analytical nature of Late Latin and Early Romance could extend it to new contexts. Por, resulting from a coalescence of Latin pro, �for this reason, because�, and per, �for the sake of, on account of, with a view to�, retains the similar purpose functions of the two Latin prepositions, and in Old Portuguese could be used in this capacity either with a noun, e g., por ele, �for him�, or with an infinitive: por defender, �for the purpose of defending�. Pera is simply the combination of por/per with a < Latin ad (Penny 1991: 196), with no change in meaning in Old Portuguese. Once infinitives began to occur with prepositions, the Latin imperfect subjunctive forms that were reanalyzed as inflected infinitives must have occurred postprepositionally as well, thereby producing the preposition-plus-inflected infinitive sequence seen in Old Portuguese. Furthermore, the analytical nature of Vulgar Latin and the Romance languages contributed to this pairing of preposition plus inflected infinitive. In the same way that Romam, �to Rome�, became ad Roma(m), and mecum, �with me�, developed to cum + mecum > Port. comigo, it would be a logical step to extend this pattern to the newly reanalyzed inflected infinitive by augmenting the purpose function with pera/por, as in the following hypothetical example: fazerem, �for them to do�-> pera fazerem.
This same process may be seen in the development of the purpose conjunction para que, �in order that�. In finite clauses, the original purpose and result conjunction in Latin, ut, is eventually replaced in Vulgar Latin and Romance by quod plus the subjunctive in purpose clauses and quod plus the indicative in result clauses (Harris 1978: 233, Bourciez 1946: 126-27). Subsequently, quod develops to que, and is then made more forceful through the addition of a preposition. For example, para is added to produce para que in purpose clauses (Harris 1978: 233). In a similar fashion, the pairing of purpose preposition with the inflected infinitive, as well as with the uninflected infinitive, may be the product of the extension of infinitive usage to postprepositional position along with the general tendency towards analytical constructions characteristic of Romance language development. The reason that the inflected infinitive is paired with pera/por as opposed to other prepositions is because it is the inflected infinitive that retains the purpose function due to its development from the Latin imperfect subjunctive. Finally, the reason that the same process does not occur with result clauses is because the subjunctive is eventually replaced by the indicative in these clauses in Vulgar Latin (Harris 1978: 233).
Thus, the pairing of both inflected and uninflected purpose infinitives with pera/por produces the following Vulgar Latin constructions from Luso-Romance territory:
|-8||Tantu nobis et vobis bene complacuit; et de precio apud vos nichil remansit per dare|
|�And both parties were in agreement on this matter; and as to your obligations of payment, nothing remained to be paid. (Leite de Vasconcelos 1922: 13)�|
|et intrarunt in placito testimoniale pro in tercio die darent testes... et Tradiderunt placitum ad legem et pervenerunt in tercio die pro ire ad legem...|
|�and they began the hearing in order to, on the third day, provide witnesses... and they put the agreement into law, and arrived on the third day in order to employ the law...� (Silveira Bueno 1968: 31)|
per dare and pro ire are purpose constructions with uninflected infinitives, and pro darent is an inflected infinitive construction. As Rodrigues points out, darent clearly shows imperfect subjunctive morphology,  but that it must be an inflected infinitive because it occurs after a preposition (1932: 4).
From the point at which pera/por joins with the inflected infinitive, it was then possible for this preposition-plus-infinitive construction to extend analogically the inflections to other postprepositional infinitive constructions, first to de, and then to other prepositions such as sem, �without�, and até, �until�. From there it continued on to non-prepositional infinitive constructions, where the infinitive is the complement of a verb: Receio terem vindo cedo demais, �I'm afraid they've come too early�. The proposed sequence of developments is the following:
1) Loss of ut plus phonological changes makes the imperfect subjunctive look like an inflected infinitive syntactically and morphologically.
2) Use of this form in purpose clauses coincides with use of the uninflected infinitive in purpose clauses.
3) interchangeability of both purpose verb forms allows uninflected purpose infinitives to adopt the imperfect subjunctive's morphology, and it strengthens the imperfect subjunctive's identity as an infinitive form.
4) These purpose infinitives (both inflected and uninflected) are incorporated into the preposition-plus-infinitive pattern through the addition of the prepositions pera/por, in line with the general Romance tendency toward analytical structures.
5) From the original purpose contexts, the inflection of the infinitive spread to other postprepositional contexts, most notably after de, and finally to the non-prepositional contexts of Modern Portuguese.
Of the two main theories on the origin of the Portuguese inflected infinitive, the imperfect subjunctive approach is the more promising because it meets morphological as well as syntactic criteria that the creative theory cannot. However, instead of deriving the inflected infinitive from the imperfect subjunctive in volitional clauses, this article has argued in favor of its origin in the Latin imperfect subjunctive in purpose clauses. This switch from volitional to purpose contexts allows the consolidation of several independent issues, including infinitive constructions, purpose clauses, and prepositional constructions, which otherwise cannot be explained.
The circumstances surrounding the origin of the inflected infinitive mandate that any analysis, including the present one, must postulate hypothetical steps that are not supported by empirical evidence. With a construction like pro darent, �in order to provide�, the origin in the imperfect subjunctive and the eventual appearance of an inflected infinitive after a purpose preposition are attested. It is the intermediate steps between these stages that remain hypothetical. Although the present article cannot improve on the hypothetical nature of this area of research, it has at least shown that purpose clauses must have played an integral role in the origin and development of the inflected infinitive in Portuguese. 
Bennett, Charles E. 1955. New Latin Grammar. Boston: Allyn.
Bourciez, Édouard. 1946. Éléments de linguistique romane. Paris: Klincksieck.
Carter, Henry H. 1967. The Portuguese Book of Joseph of Arimathea. Chapel Hill: U of North Carolina P.
Ernout, Alfred, and Francois Thomas. 1953. Sintaxe latine. 2nd ed. Paris: Klincksieck.
Gamillscheg, Ernst. 1970. Studien zur Vorgeschichte einer romanischen Tempuslehre. 2nd ed. Tübingen: Tübinger Beiträge zur Linguistik.
Grandgent, Charles H. 1907. An Introduction to Vulgar Latin. Boston: Heath.
Harrington, K.P. 1962. Medieval Latin. Chicago: U of Chicago P.
Harris, Martin. 1978. The Evolution of French Syntax. A Comparative Approach. London: Longman.
Jones, Michael. 1988. �Sardinian�. The Romance Languages. Eds. Martin Harris and Nigel Vincent. New York: Oxford UP. 314-50.
Koike, Dale A. 1983. �The Inflected Infinitive in Colloquial Brazilian Portuguese as a Marker of Social Dialects�. Spanish and Portuguese in Social Context. Eds. John J. Bergen and Garland D. Bills. Washington, D.C.: Georgetown UP. 92-104.
Leite de Vasconcelos, José. 1922. Textos Arcaicos. Lisboa: Livraria Clásica.
Martin, John W. 1960. �Remarks on the Origin of the Portuguese Inflected Infinitive�. Word l6: 337-43.
Maurer, Theodoro Henrique, 1951. Dois Problemas da Língua Portuguesa: O Infinitivo Pessoal e o Pronome SE. São Paulo: U de Sao Paulo.
_____. 1968. O Infinitivo Flexionado Português: Estudo Histórico-Descritivo. São Paulo: Editora Nacional.
Moffatt, L.G. 1967. �Origin and Development of the Personal Infinitive in Portuguese�. Revista de estudios hispánicos. 1. 1: 35-60.
Osborne, Bruce. 1982. �On the Origin of the Portuguese Inflected Infinitive�. Papers from the 5th International Conference on Historical Linguistics. Ed. Anders Ahlqvist. Amsterdam: John Benjamins. 243-48.
Penny, Ralph. 1991. A History of the Spanish Language. Cambridge: Cambridge UP.
Raposo, Eduardo. 1987. �Case Theory and INFL to Comp: The Inflected Infinitive in European Portuguese�. Linguistic Inquiry 18.1: 85-109.
Roberts, Kimberley S. 1952. An Anthology of Old Portuguese. Lisbon: Livraria Portugal.
Rodrigues, Jose María. 1932. �Sobre o uso do infinito impessoal e do pessoal em Os Lusiadas� Boletím de Filología 1: 3-7, 177-84.
Silveira Bueno, Francisco da. 1968. Antologia Arcaica. São Paulo: Editora Saraiva.
Thomas, Earl. 1969. The Syntax of Spoken Brazilian  Portuguese, Nashville: Vanderbilt UP.
Väänänen, Veikko. 1988. Introducción al latín vulgar. 3a ed. Madrid: Gredos.
Williams, Edwin B. 1962. From Latin to Portuguese. 2nd ed. Philadelphia: U of Pennsylvania P.