TRADffiONAL THEORY ABOUT THE ORIGIN OF THE LATIN IMPERFECT *

Introduction The imperfect tense expresses a continuous past action which is unfinished, as the name itself indicates im-perfectum. This characteristic accounts for its three uses: a) pure, durative imperfect b) iterative imperfect and c) imperfect de conatu. These uses are best preserved in Greek but were also used in Latin, where the forms of the old imperfect disappeared. In Proto-Germanic, the IE imperfect, the aorist and the perfect continue partly in the old perfect and partly in its counterpart, the preterite, while, in Proto-Slavonic, the old imperfect for nonmomentary actions was replaced by forms ending in *-ahb. l In ltalic languages, the functions ofthe IE imperfect passed on into the compounds with *bh7Jam.


Introduction
The imperfect tense expresses a continuous past action which is unfinished, as the name itself indicatesim-perfectum.This characteristic accounts for its three uses: a) pure, durative imperfect b) iterative imperfect and c) imperfect de conatu.These uses are best preserved in Greek but were also used in Latin, where the forms of the old imperfect disappeared.In Proto-Germanic, the IE imperfect, the aorist and the perfect continue partly in the old perfect and partly in its counterpart, the preterite, while, in Proto-Slavonic, the old imperfect for nonmomentary actions was replaced by forms ending in *-ahb.l In ltalic languages, the functions ofthe IE imperfect passed on into the compounds with *bh7Jam.

Towards the formation of the Latin im.perfect
Latin imperfect, formed with the suffix -ba-, which is a constituent part of all Latin verbs in the indicative mood except for the verb esse, functionally corresponds to the IE imperfect.It continues the imperfective aspect of the IE imperfect, which expresses a past action in progress (as unfinished), however, its use is much narrower than in Greek, for example.It is formed from the present stem of the verb by adding the suffix -ba-(in the 1 st and 2nct conjugations) or -eba-(in the Jfd and 4th conjugations); the endings are the same as for the present, except in the pt sg., where the ending is -m, which is an IE secondary ending.
The article is an abridged version of a part of the author's MA thesis on The Imperfect in Cicero's Texts, which deals with the development of the Imperfect tense from Indoeuropean to Latin as well as with its use from the beginnings of Latin literature to the period of late Latin, with particular regard to Cicero's texts.The thesis was presented on March 24 2000 before a panel of examiners consisting of the following members: Prof. Erika Mihevc Gabrovec, Prof. Matjaž Babič, and Dr. Marko Marinčič.
Cf. Brugmann 1922: 573-574, Krahe 1972: 125-126, Szemerenyi 19904: 323.The main part and basic characteristic of the Latin imperfect is the suffix -bam, which most probably originates from a form of the auxiliary verb "to be".Most grammarians agree2 that the suffix -bam deri ves from an older form * bhy-a-m, which in turn derives from the IE root *bheyH-, the meaning of which was "to grow" (hence "to originate", "to become", "to exist", "to be", "to retain", "to live").In Proto-Italic, the form * bhy-a-m developed into *fam (the proof of this can be found in the Oscan form fufans3 ); what has been left of it are -b-and -f in the imperfect of Italic languages.The Latin imperfect containing -ba-is not formally connected with the IE imperfect, but is a Proto-Italic new form and is, together with b-future, an Italic-Celtic particularity.In grammar books and treatises, the imperfect and the future are always discussed together as b-tenses.The proof that this is a particularity of Italic can be found in other Italic languages, like the aforementioned Oscan form of the 3rct pl.fufans, which functionally, but not formally, corresponds to the Latin erant; its Latin counterpart would be *fubant.Nevertheless, this is not the only prooffor b-(or f-) tenses in Italic languages.Besides the Oscan imperfectfufans, we can also find the future form carefo in Faliscan, which corresponds to the Latin form carebo and pafo or pipafo (=Lat.bibam); both forms have been preserved in inscription on goblets. 4he -b-/-f 5 future can also be found in Celtic, but not in Oscan (which hasj-perfect) or Umbrian.So, b-or j-future can be found in Celtic languages, Latin and Faliscan, while the imperfect can be found in Latin, Oscan and Umbrian.This fact indicates an entirely Italic formation.

The time of the formation of the Latin imperfect
The origin of the Latin imperfect can only be loosely defined.Due to the Oscan formfufans (= Lat.erant) "they were" (the only preserved form of this imperfect in Oscan) formed from *bhu-bhyant, it is presumed that this periphrastic formation was already present in Proto-ltalic.However, one needs to be careful when discussing the imperfect as a Proto-Italic formation.Namely, the stem of the aforementioned Oscan imperfect is questionable, as there only exists one example, and also because the formation of the imperfect from the root *bhu-with the suffix *bhl}-a-derived from the same root, is not highly likely; Latin also did not form the corresponding *fubant, but favoured erant 6 instead, although, by its formation, it falls out of the frame of other b-imperfect tenses. 7 The question of the chronological origin of the future and the imperfect still remains open.As Old Irish forms thef/b-future from derived verbs, a conclusion was made long ago about a common origin of the future form in Irish, Latin and Faliscan.This could have happened in the area to the north of the Alps before the Latin peoples settled in Italy, in the times when the Irish and Latin predecessors were neighbours.The connection between the Latin and Old Irish future forms was one of the most important points on which the theory about the existence of a proto-Latin-Irish community was based (Walde 1917).As the Oscans and Umbrians formed the future with the suffix -s-, and not the suffix -b-, it can be assumed that, in the times of the formati on of the future forms, they were not yet neighbours to the Latin peoples; on their arrival to the Italian Peninsula, when they settled near the Latin, they already had an existing future form.The future is therefore older than the imperfect; the latter was only formed in Italy from where it penetrated into Oscan.The origin of the future form dates back into the times when Celtic, Latin and Faliscan were still closely connected, while the origin of the imperfect dates back into the times of closer connection between Latin, Oscan and Umbrian (Pohl 1986: 208;Walde 1917).Leumann believes that the imperfect is older than the future because it appears in all Latin verbs (with the exception of esse) and can also be found in Celtic.The -be/ofuture is more recent since, in Latin, it cannot be found in inherited thematic primary verbs (most ofwhich belong to the 3fd and 4th conjugations) which used the old e-future.At the same tirne, he advocates the necessary distinction between the imperfect and the future, as both are unequally divided not only by languages, but also within one language into paradigms and verb classes.(Leumann 1924: 60-75, LHS 19775: 579).

Traditional theory about the origin of the Latin imperfect
Traditional theory explaining the origin of b-tenses is "Kompositionstheorie" (composition theory).According to this theory, the auxiliary verb *bheljH-"to be" is added to the pure verb stem or the nominal form.When forming the future of the 1 st, 2nct and partly 4th conjugations, we add the morpheme -be/oto the last syllable of the present stem and conjugate it like we do the present indica ti ve of verbs in the 6 The suffix -ii-in eram, eriimus undoubtedly originates in the IE proto-language in some aorist formation (Cf: Old Irish.bel "I was" (< *bhyiim), Lithuanian.biwo "he was" (< *bhyiit)).However, we cannot prove the aorist with the root *es-in any IE language; hence the uncertainty regarding the origin of the forms eram, eriis ... cf.Safarewicz 1969, 226-228. 7 Cf.also reduplication in the Oscan perfect.fufens"fuerunt" and the Umbrian future ex.fefure; for a more detailed explanation see Planta 1892/1897: 2, 373; 2, 3282; 2, 331; 2, 342, Buck 1904:  § 128, 2a,  § 193,   Brugmann 1897/19162: II2, 3, 508  § 421, 7 A.2; II2, 3, 506, Leumann 1924: 66-68, Hermann 1948, LHS   1977 5 : The present stem also appears in esse and ire: esse The common feature of all explanations of the Latin imperfect is that they look for the old preterite form of the stem *Ju-, from which Latinfuf and Old Latin conjunctive fuiim are derived ( cf. also of Old Indian bhuand Greek epu-"to form, to become, to be"), in the ending -ba-m; this was formed from *-bhy-ii-m and corresponds in its formation to er-ii-m (< *es-ii-m), which was derived from the stem *es-.By its origin, the imperfect form is therefore a periphrasis created with the inflected form of the verb "to be".
On the other hand, explanations of the stem part of the imperfect form differ, because the structure of the part of the verb before the suffix remains unknown.Some grammarians see in it a stem incomprehensible to us today; others a flexible form that was still alive in Latin in historic times, but which later underwent such changes that it cannot be recognised anymore.The fusion oftwo stems was supposed to correspond to the development of the Romance future tense from the Proto-Romance infinitive + *habyo (e.g.*cantdre hdbyo > Fr. chanterai, It. cantaro; Sihler 1995: 554-555).The l5 1 , 2nct and partly 4th conjugations of dare, žre and esse have the present stem while, in the 3rct and 4th conjugations (cole-bam, lenie-bam, capie-bam), the present stem is somewhat remodelled; namely, cole-, lenie-, capie-only appear in this form and, therefore, cannot be regarded as established variants of the verb stem; between the stem and the morpheme there is -e-, for which there is still no adequate explanation.Efforts to discover the origin ofthe stem of the yct and 4th conjugations led to several hypotheses.

Attempts to explain the verb stems of the 3ro and 4th conjugations
The first part of the imperfect, ending in -e, could have derived from the 3rct conjugation, and probably became equal to the first part of the Slavonic imperfect in -e ( cf.Lat.vehe-bam = Old Church Slavonic veze-acho ); it was similar to a case of a nominal formation or a verbal noun.Later on, forms with -iebam in the Jrd and 4th conjugations probably appeared.However, if this first part is a case of a verbal noun ending in -e, a question arises how the corresponding verbs ending in -a, -e and -1 in the ist, 2nct and partly 4th conjugations were formed.Bopp placed the Latin imperfect side by side with the Slavonic imperfect (Bopp 1833/1849: n2 399ss.).Schmidt (Schmidt 1871), too, directly equated Latin lege-in lege-bam with Slavonic nese-in nese-acho.Both authors influenced subsequent research (cf.Brugmann 1897 /1916 2 : II 3 2 , 506).In this case, we see in lege-either a) a pure verb stem without an inflexional ending, b) a stem of a verbal noun, or c) a fossilised case of a verbal noun without the endingcasus indefinitus (Hoffmann 1920(Hoffmann /1924)).The weak point of this explanation is the assumption that this is a form which cannot be proved to be an independent and living verb form in Latin.
Hermann (Hermann 1951) tried to prove that the first part of the compound is neither a verbal noun nor casus indefinitus, but a pure stem preserved in combination with * bhl}am.He explains the long e in the first part on an example of primary verbs ofthe 2nct conjugation (e.g.ple-bam, sile-bam,fide-bam ... ), from where it was to spread to derivatives (e.g.albe-bam,flore-bam, noce-bam ... ).He rejected the attempt to explain the long e with an ad hoc invented participle ending in -e. 8ommer's objection that the imperfect, ending in -bam, is an ltalic new form that cannot be directly linked to any IE proto-form (Sommer 1914: 140ss.) was rebutted by Giintert (Giintert 1917).According to the latter, the imperfect lege-bam is an adapted form of an older pre-ltalic IE verb form: Giintert saw in lege-the stem of the Greek root aorist ofthe type EAE'(lJ.9He considered the ltalic imperfect asa continuation ofthe IE aorist ofheavy bases; thus, the original *lege-t (= Gr.EAEy"l]('t")) still present in ltalic was supposed to be replaced by a new formation lege-ba-t.He categorically defended the opinion that *-fam was originally an independent auxiliary verb directly added to the old aorist stem; he tried to prove with Celtic and Latin compounds ( e.g.calefacio, liquefacio) that adding an auxiliary verb to a verb s tem in ltalic was nothing unusual.The weakness of his attempt to interpret the imperfect is that he was forced to base the formation of the imperfect entirely on the imperfect forms of the Jrd conjugation, as it is only here that the stem in -e can be found in pure form; from here it was to spread by analogy to other forms within the conjugation and on to other conjugations.Evidently, Giintert realised the weakness of his attempts himself.Namely, despite advocating lege-as an aorist stem, he soon came up with a surmise that lege-in lege-bam might originally not be a stem after all, but rather a flexible aorist form (Giintert 1917: 18)  The deficiency of his explanation was also critically highlighted by Hoffmann (Hoffmann 1920(Hoffmann /1924)), whose two main doubts were that the Latin imperfect has an unfinished meaning and not an aorist one, and that the merging of two inflected forms into one is a rather unusual phenomenon.
Stowasser and Skutsch took a different approach (Stowasser 1901, Skutsch 1914:  283-292).In amii-, lege-, audie-, they looked for a usual verb form preserved in Latin whose meaning would easily explain its connection with -bam.In imperfect forms they saw formations composed of a participle and an auxiliary verb.The result was supposed to be the following: * audients bhljiim This theory was widely approved by classical philologists, yet it left many questions unanswered.The first problem is that, in Old Latin, common imperfect forms ofthe 4th conjugation were not audie-bam (as would be expected ifthe base form was *audiens-bam), but audz-bam.However, audz-bam could be a more recent formation than audie-bam and could first have been remodelled in the vernacular by analogy with amii-re:amii-bam, dele-re:dele-bam. 10Furthermore, there is still no satisfactory 10 The forms of the imperfect of the 4th conjugation in -fbam appear throughout Latin literature, in archaic period mainly and in c!assi~al period only with poets (as an archaism and a metric aid), while prose writers avoid it.There is no irrefutable evidence as to which forms are o!der, as there is no form in -fbam oriebam.A possible explanation is that forms in -fbam appeared by an analogy because, in conjugations with the base containing a long vowel (1' 1 , 2nct and 4 1 h), the language saw a certain whole in comparison with the yct conjugation, which was left out by this analogy; in this case, the forms in -ibam are probab!yyounger.On the basis of the repetition of forms in O!d Latin, we cannot determine whicii.form is older; forms in -fbam or -iebam appear side by side, and there are too few prose works preserved for comparison.
explanation why the imperfect form of the verb žre did not preserve the anticipated form *ie-bam "I went" (from *iens-bam), but was replaced by a new formation f-bam, while the form audie-bam was preserved side by side with audf-bam.11 The expected phonological development also throws doubts on the derivation of lege-bam from * legens-bam.According to phonological laws, the cluster -nsf-could not develop into -b-, which is characteristic of the imperfect.Expected development 1: The only answer would be that n before s disappeared early and that -sf-changed into -fvia 1!-(e.g.. difficilis from *disfacilis), and further into -b-.Expected development 2: Although generally accepted, Stowasser and Skutsch's hypothesis did not hold water, so the search for a living Latin verb form which could serve as the base for the derivation of the imperfect continued.One such verb was pointed out by Lowe, who saw in the stem of the German weak preterite (salbo-da) some shortened infinitive composed with the verb "to do", and incidentally remarked that this was probably the way the Italic imperfect had been formed. 13According to him, imperfect forms originated with mechanical clipping, which is how *amiire-bam turned into amii-bam.Yet, the theory does not hold in the 3rd conjugation, as it fails to explain how the form *legere-bam turned into lege-bam.In the Jrd conjugation, Lindsay anticipated the working of an analogy (Lindsay 1897: 563-565).Amii-, vide-,finž-can be treated as pure verb stems, which does not hold true for lege-, because originally, its verb stem was lege-.By analogy, verbs of the 3rct conjugation thus probably followed verbs ofthe 2nct conjugation; the origin ofthe form lege-bam could have followed the example of vide-bam with the analogous transfer of the long vowel.In verbs of the 4th conjugation, such change in the formation ofthe imperfect probably occurred in the 2nct century BC.
Hoffmann took an infinitive originating in the locative of a verbal noun as the starting point for his theory (Hoffmann 1920/24, 222).Thus the infinitives parii-re, lege-re originate in *parii-se, *lege-se and these two forms (presumably) in *para-si, *lege-si respectively.Since the infinitive is the locative by origin, we can presume its original locative meaning, if linking the infinitive with the past form *1am14 • This would mean that the Latin imperfective past was described: *pariisifiim "I was at preparing" = I was preparing *legesifiim "I was at reading" = I was reading.
Sommer, too, sees in age-bam, like in compounds of the type calefacio, an infinitive formation, and points out the parallel with the Slavonic imperfect *nese-achb > neso, which has been explained with the fusion of such infinitive with *es-o-m "I was": *nese-esom "I was at carrying" the same as age-bam "I was at leading" (Sommer 1948: 521).
The explanation for the stem of the verbs of the 3rd and 4th conjugations was also sought in adverbs 1-licet, vide-licet, sc1-licet, in verbs of the type calefacio and in impersonal verbs.Adverbs 1-licet, vide-licet, scf-licet are undoubtedly compounds with infinitives (žre, videre and scfre).The question arises about the occurrence of fusion.Presumably, it is younger than the imperfect forms 1-bam, vide-bam and it cannot be said with certainty whether the infinitive forms were still *1se, *videse, *sc1se, or already žre, vid re, *scfre.Expected development: a) the short e in compounds disappears b) s (or r) assimilates into l c) -II-after a long vowel changes into -/-.*1se-licet According to this pattern, the fusion of the infinitive and the auxiliary into the imperfect form would also be possible (Lindsay 1897: 563-565, Hoffmanu 1920/1924: 229-230, LHS 19775: 566).
16 Replacement of -s-, which disappeared before a voiced consonant, by lengthening the previous vowel can also be found in historic limes, e.g.comis < Old Latin cosmis (Duenos), diimus < Old Latin dusmos (Liv. Andr. trag.39 dusmo in loco) Hoffmann 17 anticipates a similar process with verbs of the type cale-facio, which are mostly bound to the 2nct conjugation by their origin (Hoffmann 1920(Hoffmann /1924: 230-231: 230-231).They probably developed from the connection of the infinitive + facio: calefacio < *calesefacio like vide-licet from *videse licet.A partial confirmation of such a supposition can be found in Old Latin, where we can find the connection of the verb facio with AcJ18, e.g.
Lucil.1270 purpureamque uvamfacit albam pampinum habere Varro rust.3, 5, 3 quod earum aspectus ac desiderium marcescerefacit volucres inclusas Phonologically, too, calefacio can be derived from *calesefacio without any major problems.Expected development: a) the vowel in the unaccented mid-syllable position disappears b) -s-assimilates in to 1c) -jJ-after a long vowel changes into 1- The latter statement triggers the questions of the appearance of these forms a) in tmesis, e.g.Cato agr.157, 9 ferve bene facito (as opposed to forms with no tmesis 1 9); Varro rust.The relationship between cale-in calefacio and cale-in the imperfect cale-bam remains a matter for discussion.If there had been a connection between them, it would have to be preserved through the history of the language.So, even after the appearance ofthe imperfect (*calefiim, *sanafiim), the forms cale, sana should have remained independent infinitives, but were not preserved in the language at all.The hypothesis that the stem part with -eoriginates in impersonal verbs of the 2nct conjugation asa verbal noun with-e-(e.g.*taede bat "there was disgust" ="it was disgusting"), has not been widely accepted.20 17 Hoffmann 1920/1924: 230-231. 18 Cf. Schmalz 1928: 426. 19  The last hypothesis regarding -eis offered by Matasovic (Matasovic 1997: 220-221): the element -ebefore the suffix -hain the verbs of the 3rct and 4th conjugations might have originated in IE suffix -eh r, which was used for the formation of durative verbs expressing state; namely, a similar suffix also appears in the non-terminative past tense in Slavonic languages, for example Old Church Slavonic gre-

beahb.
* What all traditional hypotheses have in common is that they consider imperfect forms as compounds, as descriptive combinations with the auxiliary *jam (<*bhl}am) "I was".However, none of the hypotheses presented has provided a satisfactory explanation of the origin of the imperfect.More acceptable and also more plausible answers are offered by the modem theory about the origin of the imperfect: it explains the origin of the imperfect forms with the transition of IE verb categories into Latin verb categories and with interna!adjustment and organisation of these.
579.   3rct conjugation.With the imperfect, we add the morpheme -ba-to the last syllable of the present stem and conjugate the new form the way we conjugate the present indicative, with the exception of the l5 1 sg., which has the IE secondary ending -m.