and if it were possible to separate in every instance these cases from those in which the imperfect may be said to have been required, we should have a criterion by which we might distinguish this use of the imperfect from others. But since the progressive function of the tense is not altered, such a distinction is not necessary.

Statistics as to the frequency of the imperfect of situation in early Latin are worth little because the chief remains of the language of that period are the dramatists in whom naturally the present is more important than the past. The historians, to whom we should look for the best illustrations of this usage, are for the most part preserved to us in brief fragments. Nevertheless an examination of the comparatively few descriptive passages in early Latin reveals several points of interest.

In Plautus and Terence the imperfect was not a favorite tense in descriptions. Bacch. 258-307, a long descriptive passage of nearly 50 lines, interrupted by unimportant questions, shows only 4 imperfects (1 aoristic) amid over 40 perfects, historical presents, &c. Capt. 497-515, Amph. 203-261, Bacch. 947-970, show but one case each. Stich. 539-554 shows 5 cases of erat. In Epid. 207-253 there are 10 cases.

In the descriptive passages of Terence the imperfect is still far from being a favorite tense, though relatively more common than in Plautus, cf. Andr. 48 ff., 74–102, Phorm. 65-135 (containing 11 imperfects). But Eunuch. 564-608 has only 4 and Heaut. 96-150 only 3.

Another very instructive passage is the well-known description by Q. Claudius Quadrigarius of the combat between Manlius and a Gaul (Peter, Hist. rom. fragg., p. 137, 10b). In this passage of 28 lines there are but 2 imperfects. The very similar passage describing the combat between Valerius and a Gaul and cited by Gellius (IX, 11) probably from the same Quadrigarius contains 8 imperfects in 24 lines. Since Gellius is obviously retelling the second story, the presumption is that the passage in its original form was similar in the matter of tenses to the passage about Manlius. In other words Gellius has 'edited' the story of Valerius, and one of his improvements consists in enlivening the tenses a bit. He describes the Manlius passage thus: Q. Claudius primo annalium purissime atque illustrissime simplicique et incompta orationis antiquae suavitate descripsit. This simplex et incompta suavitas is due in large measure to the fact that

Quadrigarius has used the simple perfect (19 times), varying it with but few (4) presents and imperfects (2). A closer comparison of the passage with the story of Valerius reveals the difference still more clearly. Quadrigarius uses (not counting subordinate clauses) 19 perfects, 4 presents, 2 imperfects; Gellius, 4 perfects, 9 presents, 8 imperfects. In several instances the same act is expressed by each with a different tense:

Gellius has secured greater vividness at the expense of simplicity and directness.

This choice of tenses was, as has been said, a matter of art, whether conscious or unconscious. The earlier writers seem to have preferred on the whole the barer, simpler perfect even in passages which might seem to be especially adapted to the imperfect, historical present, &c. The perfect, of course, always remained far the commoner tense in narrative, and instances are not lacking in later times of passages' in which there is a striking preponderance of perfects. Nevertheless the imperfect, as the language developed, with the growth of the rhetorical tendency and a consequent desire for variety in artistic prose and poetry, seems to have come more and more into vogue.'

The fact that the function of a tense is often revealed, defined, and strengthened by the presence in the context of particles of various kinds, subordinate clauses, ablative absolutes, &c., &c.,

1 E. g. Caesar, B. G. I. 55 and 124-5.

'The relative infrequency of the tense in early Latin was pointed out on p. 164. Its growth as a help in artistic prose is further proved by the fact that the fragments of the later and more rhetorical annalists, e. g. Quadrigarius, Sisenna, Tubero, show relatively many more cases than the earliest annalists. This is probably not accident. When compared with the history of the same phenomenon in Greek, where the imperfect, so common in Homer, gave way to the aorist, this increase in use in Latin may be viewed as a revival of a usage popular in Indo-European times. Cf. p. 185, n. 2.

3

was pointed out in Trans. Am. Philol. Ass. XXX, pp. 17 ff. What was there' said of Plautus and Terence may here be extended to the whole period of early Latin. The words and phrases used in this way are chiefly temporal. Some of those occurring most frequently are: modo, commodum; tum, tunc; simul; dudum, iam dudum; iam, primo, primulum; nunc; ilico; olim, quondam; semper, saepe; fere, plerumque; ita,' &c., &c. A rough count shows in this class about 120 cases, accompanied by one or more particles or expressions of this sort. Some merely date the tense, e. g., tum, modo, dudum, &c. Others, as saepe, fere, primulum, have a more intimate connection with the function. Naturally the effect of the latter group is clearest in the imperfects of customary past action, the frequentative, &c., and will be illustrated under those headings. Here I will notice only a few cases with iam, primulum, &c., which illustrate very well how close the relation between particle and tense may be. The most striking cases are:

"

Plautus, Merc. 43, amare valide coepi[t] hic meretricem. ilico

Res exulatum ad illam <c>lam abibat patris. Cf. Men. 1116, nam tunc dentes mihi cadebant primulum.

id. Merc. 197, Equidem me iam censebam esse in terra atque in tuto loco:

Verum video ...

id. Cist. 566, Iam perducebam illam ad me suadela mea, Anus ei <quom> amplexast genua ...

id. Merc. 212, credet hercle: nam credebat iam mihi.

The unquestionably inceptive force of these cases arises from the combination of tense and particle. No inceptive' function can be proved for the tense alone, for I find no cases with inceptive force unaccompanied by such a particle.

'Cf. also Morris, Syntax, p. 83.

How far the nature of the clause in which it stands may influence the choice of a tense is a question needing investigation. That causal, explanatory, characterizing, and other similar clauses very often seem to require an imperfect is beyond question, but the proportion of imperfects to other tenses in such clauses is unknown. Cf. p. 166, n. 1.

No introductory conjunctions are included in this total, nor are other particles included, unless they are in immediate connection with the tense.

In Trans. Am. Philolog. Ass. XXX, p. 21, I was inclined to take at least Merc. 43 as inceptive. This I now believe to have been an error. The inceptive idea was most commonly expressed by coepi+infin. which is very common in Plautus and Varro. We have here the opposite of the phenomenon discussed on p. 177.

There are a few cases in which the imperfect produces the same effect as the imperfect of the so-called first periphrastic conjugation: Terence, Hec. 172, Interea in Imbro moritur cognatus

senex.

Horunc: ea ad hos redibat lege hereditas.=reditura erat, English 'was coming', 'was about to revert', cf. Greek μed with infinitive.

Cf. Phorm. 929, Nam non est aequum me propter vos decipi, Quom ego vostri honoris causa repudium alterae

Remiserim, quae dotis tantundem dabat.=datura erat &c. In these cases the really future event is conceived very vividly as already being realized.

Plautus, Amph. 597 seems to have the effect of the English 'could':

Neque ... mihi credebam primo mihimet Sosiae

Donec Sosia . . . ille . . .

But the 'could' is probably inference from what is a very vivid statement. A Roman would probably not have felt such a shading.1

I B. THE IMPERFECT OF CUSTOMARY PAST ACTION. The imperfect may indicate some act or state at some appreciable distance in the past as customary, usual, habitual &c. The act or state must be at some appreciable distance in the past (and is usually at a great distance) because this function of the tense depends upon the contrast between past and present, a contrast so important that in a large proportion of the cases it is enforced by the use of particles. The act (or state) is conceived as repeated at longer or shorter intervals, for an act does not become customary until it has been repeated. This customary act usually takes place also as a result or necessary concomitant of certain conditions expressed or implied in the context, e. g. maiores nostri olim &c., prepares us for a statement of what they used to do. The act may indeed be conceived as occurring only as a result of a certain expressed condition, e. g. Plautus, Men. 484 mulier quidquid dixerat,

Some of the grammars recognize 'could' as a translation, e. g., A. & G. § 277 g.

E. g. tum, tunc, olim &c. with the imperfect, and nunc &c. with the contrasted present.

Idem ego dicebam my words would be uttered only as a result of hers.1

There are 462 cases of the customary past usage of which 218 occur in independent sentences, 244 in dependent. This large total, more than one-third of all the cases, is due to the character of Varro's De lingua latina from which 289 cases come. This is veritably a 'customary past' treatise, for it is for the most part a discussion of the customs of the old Romans in matters pertaining to speech. Accordingly nearly all the imperfects fall under this head. Plautus and Terence furnish 112. The remaining 61 are pretty well scattered.

As illustrations of this usage I will cite (arranging the cases according to the classes of verbs):

I. I. Plautus, Pseud. 1180, Noctu in vigiliam quando ibat miles, quom tu ibas simul,

Conveniebatne in vaginam tuam machaera militis?

Terence, Hec. 157, Ph. Quid? interea ibatne ad Bacchidem? Pa. Cottidie.

Varro, L. L. 5. 180, qui iudicio vicerat, suum sacramentum e sacro auferebat, victi ad aerarium redibat.

I. 2. Plautus, Bacch. 429, Saliendo sese exercebant magis quam scorto aut saviis. (cf. the whole passage).

Hist. fragg., p. 83. 27, Cn., inquit, Flavius, patre libertino natus, scriptum faciebat (occupation) isque in eo tempore aedili curuli apparebat, .

I. 3. Terence, Eun. 398, Vel rex semper maxumas

Mihi agebat quidquid feceram :

Varro, L. L., 5. 121, Mensa vinaria rotunda nominabatur Cilibantum ut etiam nunc in castris. Cf. L. L. 7. 36, appellabant, 5. 118, 5. 167 &c.

1 This usage seemed to me formerly sufficiently distinct to deserve a special class and the name occasional', since it is occasioned by another act. It is at best, however, only a sub-class of the customary past usage and in the present paper I have not distinguished it in the tables. It is noteworthy that the act is here at its minimum as regards repetition and that it may occur in the immediate past, cf. Rud. 1226, whereas the customary past usage in its pure form is never used of the immediate past. The usages may be approximately distinguished in English by used to', 'were in the habit of' &c. (pure customary past), and 'would' (occasional), although 'would' is often a good rendering of the pure customary past. Good cases of the occasional usage are: Plautus, Merc. 216, 217; Poen. 478 ff; Terence, Hec. 804; Hist. fragg. p. 202. 9 (5 cases), ibid. p. 66. 128 (4 cases).