¹ Cf. Neue-Wagener, Lateinische Formenlehre 3.450 ff.

² See Caix, GFR 1.229 ff. (1878); d'Ovidio, GFR 2.63 ff. (1879); Gaspary, ZRPh. 3.622 (1879); Meyer-Lübke, ZRPh. 9.223 ff. (1886); Rom. Gramm. 2.§266; Introducción a la lingüística románica³ 190 f. (§101), 293 f. (§190); Histor. Gramm. der frz. Spr.^4–5 244 (§331); Das Altlogudoresische 46; Grandgent, Introduction to Vulgar Latin 177 ff. (§§423 ff.); Bourciez, Éléments de linguistique romane³ 82; Menéndez Pidal, Manual de gramática histórica española⁶ 310 ff.; Orígenes del español 380 f.; von Wartburg, ZRPh. 42.373 (1922); Wagner, ID 15.4 f., 14 (1939); Rosetti, Istoria limbii române² 1.133 (1940).

³ Rom. Gramm. 2.§265. Contrast the normal development of -áui- in clāuis > Ital. chiave, Span. llave; grauis > Ital., Sp. grave; nāuis > Ital., Span. nave; etc. The difficulty was perceived by Diez (Gramm. des langues romanes 2.137), who thought (following Delius) that ‘la langue populaire ... n'a fait qu'ajouter à la forme sourde [?] cantà un o comme voyelle d'appui, de même qu'elle a ajouté cette voyelle dans cantan-o [cf. also uscío, vendeo, feo = uscí + o, etc.]: cantó est donc syncopé de cantao, comme vo de vao = vado.‘ But this explanation, as Caix correctly remarks (GFR 1.229 [1878]), will not hold for Spanish, which also has cantó.

⁴ C. F. Bauer (The Latin perfect endings -ere and -erunt 22, with fn. 26; 45, §36b [Lang. Diss. No. 13, 1933]) proposes to read dormirunt in Plautus, Poen. 21, but all the manuscripts give dormierunt. In general, Bauer's work is a very careful one and most useful; but he does not consider or even cite Burger's and Meillet's articles, and adheres completely to the old idea that the short forms are ‘contractions’ of the long ones.

Neue-Wagener gives no material from Plautus, so we must complete it from Bauer 45, §36b: there are only two examples of -īuērunt (scīuērunt, Poen. 123; trīuērunt, Pseud. 818) and one of -ierunt (dormierunt, Poen. 21), which Bauer would correct to dormīrunt (see above). The verb eō and its compounds (abeō, exeō, ineō) never have any forms with (-ierunt, iere, exierunt, iniere, periere).

Cf. Burger, Études de phonologie et de morphologie latines 124 (Neuchâtel, 1928); Sommer, Handbuch der lateinischen Laut- und Formenlehre³ 566 ff.; Neue-Wagener 3.452 ff.; Meillet and Vendryes, Traité 258 ff. The ending -ierunt is preserved in Spanish (e.g. oyeron < audierunt; see, however, Baist, GGr.² 1.913; Meyer-Lübke, ZRPh. 9.234, 254) but not in Portuguese. See TLL s.v. audiō; Sommer, Handbuch³ 565 ff.; Burger, Études 124 ff., where the types audierō (Sp. oyero oyeres, etc.), audieram (Sp. oyera), audierim are also studied. See on these also Meillet, Introduction à l'étude comparative des langues indo-européennes⁷ 214; Mél. de Saussure 99; Mariès, Rev. des ét. arm. 10.169 ff. (1930).

⁵ See Leumann, Lat. Gramm. 112 ff.; Sommer, Handbuch³ 158 f.

⁶ REL 4.115 ff. (1926); Études 39 ff., 85 ff. Burger then correctly points out (REL 4.116; Études 107 f.) that the perfects lāuī, fāuī, where the belongs to the root (cf. Gk. λόϵσσα, OIr. loathar, Lat. lūstrum), never show the so-called ‘contracted’ forms; contrast nāstī: nāre.

⁷ See Meyer-Lübke, Rom. Gramm. 1.§442; Meyer-Lübke and Castro, Introducción 190 f. (§101), with Castro's footnote; Bourciez, Élém.³ 166 f. (§172); Burger, Études 96 f. The only apparent exception (pauōrem, represented by *paōrem, Ital. paore, Fr. peur) is one more case in which Romance preserved early Latin conditions, as *paōrem is precisely the regular Latin form to be expected, since (cf. sōpiō = Ved. svāpáyāmi, etc., see Burger 92 ff.).

⁸ See Burger, Études 115 ff.; Sommer, Handbuch³ 562 ff., 588.

⁹ The perfects in -īuī, -iī, -ī of some third conjugation verbs (pōnō, sapiō, etc.; see Sommer, Handbuch³ 565, 573) come under the same head as those of the verbs in -īre. For the verb eō, cf. the attestations of (-)īt, -īstī, -ī (Sommer 589); add abīsse, Miles 1197; more material in Neue-Wagener 3.466 ff. The stem of īre was of course originally ei- (i-), not ī-; iī comes probably from *iy-ai, cf. Indo-Aryan iy-āy-a (Sommer, Handbuch³ 567). The penetration of in this verb is late and rare; Plautus has only one example, amb-īuī (the usual form is iī; see Engelbrecht, WSt. 6.240 ff. [1884]); Cicero, Caesar, Sallust, and lávy never have īuī.

¹⁰ Probably for metrical reasons (-∪-); as pointed out by Engelbrecht, WSt. 6.219, Terence's use of the longer forms of the perfect only at the end of the verse, and of the shorter forms elsewhere, indicates the currency of the latter in the sermo urbanus in Terence's time. See also Meyer-Lübke, ZRPh. 9.247.

¹¹ These and the following indications also include the ‘short’ forms of the other tenses of the perfect system; but the facts are even clearer, on the whole, for the three persons of the perfect studied here (and especially for the two second persons). Thus, in Cicero we find audīstī 14 times and audīuistī once (doubtful moreover); audīstis 74 times and audīuistis once (also doubtful); see Burger, REL 4.118 (1926). The reason why cantāram, cantārō, etc. are less frequent in our texts than cantāstī, cantāsse, etc. is given by Burger, Études 119 (see below, §5).

¹² That the ‘short’ forms were the usual ones is attested for their times by Varro (Frag. 11, Goetz-Schöll² 241), Cicero (Orat. 47, 147) and Quintilian (1.6.16); see Sommer, Handbuch³ 563; Burger, REL 4.212 ff. (1926); Études 112.

¹³ Livy, who is by no means a ‘vulgar’ writer, has (according to our manuscripts) (E. B. Lease, AJP 24.415 (1903]). The form -ārunt isc omparatively more frequent in the fourth and in the fifth decades than in the first . This corresponds to the very literary and artificial character of the first decade, and is fully confirmed by the figures for -ēre (a purely literary form) as opposed to , the spoken and vulgar form, which has passed into Romance (cf. Lease, 409 ff.); the following percentages are for -ēre: 1st decade 54.7% (77.2% in Book 3), 4th decade 13.5% (4% in Book 39), 5th decade 10.1% (3.3% in Book 41).

Lucretius has only one occurrence of -āuērunt (6.3), otherwise always -ārunt, a fact which F. Muller (Mnemosyne 56.354 [1928]) correctly considers indicative of the comparatively colloquial character of Lucretius' style: ‘Porro formae, quales sunt imitāt (< -āu[i]t [sic!]) 1.70, disturbāt 6.587, luculenter demonstrant Lucretium hac in re uestigiis atque usui sermonis cotidiani quam artissime haerere.‘

¹⁴ Cf. Neue-Wagener 3.480 ff.; Sommer, Handbuch³ 564; Burger, Études 117, and see below; on the Romance forms of crēuī, mōuī, see Meyer-Lübke and Castro, Introducción³ 295 (§191), and Burger, Études 107 f., 111, 117. On lāuī, fāuī, cāuī see Priscian 2.508.6 (cf. Neue-Wagener 3.478 f.); Burger, Études 107 ff.; Sommer, Handbuch³ 562.

¹⁵ On the importance of the literary trend towards ‘regularizaron’ in the preference for the longer forms with , cf. also Engelbrecht, WSt. 6.220, and Meyer-Lübke, ZRPh 9.248 (1885): ‘. . . da wählte die Vulgärsprache die kürzere Form . . ., die klassische Sprache aber mit dem jeder Kunstsprache [!] eigenen Trieb nach höchster Regelmässigkeit (was hat nicht in solchem Streben das klassische Sanskrit geleistet!) hielt sich an die längern.‘

¹⁶ The accent of is of course contrary to the general rules of Latin accentuation; but it is well attested by the grammarians (Priscian GL 2.130.1 f.; Servius ad Aen. 3.3; cf. also 1.451) and is no more unusual than many other ‘exceptions’ to these ‘rules’ which may be found in Sommer, Handbuch³ 87 f.; Lindsay, Lat. Lang. 163 f.; Leumann, Lat. Gramm. 181 ff. (where, curiously enough, are not mentioned); cf. also Sturtevant, Pronunciation of Greek and Latin² 179. The most directly comparable accentuation is that of . The Romance languages frequently offer excellent confirmations of ‘exceptional’ Latin accentuation ; they likewise add more material for Vulgar Latin: , etc. Cf. Meyer-Lübke, Rom. Gramm. 1.§§602, 603. Furthermore, the measure - - ∪ for and similar cases in Ausonius and other late poets confirms the accent indicated by the grammarians. And we see now very well that Quintilian (1.5.22 ff.) was right in his accentuation Cámillus Céthēgus, since we know that these names are Etruscan (Schulze, Eigenn. 322).

The accent of has been shifted (or preserved?) through the analogy of all the other forms of the perfect , a phenomenon which could not take place in the present tense . In this way, a valuable distinction was established between the 3d singular present and perfect. This seems to me the only possible explanation of the Romance and Latin form , since I cannot understand why the whole form should be considered analogical instead of admitting that the accent only was shifted (or preserved) by analogy.

¹⁷ No is attested, so that the real conjugation of the perfect of fleō is flēuī flēstī (twice, flēuistī once, Lygd. 6.40) flēuit flēmus *flēstis flērunt (Verg., Ovid., Val. Fl., Statius; also flēuēre). Likewise, we find suēmus, consuēmus, but no *(con)suēuimus. See TLL s.v. fleō; Neue-Wagener 3.480. Consequently, Meillet's skepticism (BSL 27.234.6 [1927]) about the antiquity of nōmus seems excessive.

¹⁸ On nōmus, ēnarrāmus, etc., see also Engelbrecht, WSt. 6.224 ff.; on inīmus, abīmus, Verīmus (?) ibid. 235, 240; on īt ibid. 239 (and TLL s.v. eō); on perī ibid. 243.

¹⁹ The inscriptions have occurrences of the type posī (beside posīuī), perfect of pōnō; of these, two are old: posīt, CIL 1.1780; poseit 1781 (both from Rome). Plautus has imposīsse (Most. 434); Vergil has dēposīsse (Catal. 10.16, assured by the meter); see Sommer, Handbuch³ 573.

²⁰ For pōsiM, dēsī, quaesī, etc. (posīt, dēsīt, quaesīt, etc.) Burger (Études 126 ff.) considers, chiefly on the basis of the Romance forms (It. pọsi, pọse [OIt. also puosi, puose], etc.; chięsi, chięse, etc.), that the accent was on the penult.

²¹ BSL 34.127 (1933); cf. also Introduction⁷ 214. See also Pedersen, Hittitisch und die anderen indoeuropäischen Sprachen 94 ff.

²² Dīxtī, dīxtis are older than dīxistī, dīxistis; see Burger, Études 103 ff., and cf. Neue-Wagener 3.500 ff. They are frequent in archaic Latin, and normal in Terence. Burger correctly remarks (108, 133) that läuistī, mōuistī must be divided lāui-stī, mōui-stī, the roots being bisyllabic (Gk. λόϵ-σσα, etc.); the ending is then -stī, not -istī. The forms fēcistī, ēgistī, cucurristī, etc. are of course formed on the analogy of lāuī lāuistī.

²³ The final -n of Hitt. pesten is analogical, according to Petersen, who also considers the original Tocharian 2d plural ending to have been -ste. We may perhaps compare the final -ν of Gk. φέρομϵν to Lith. and Czech neseme (Vedic, Avestan -ma?) or also the Vedic 2d pl. sthána, vádathana, itana, pūnitána; see Brugmann, Grundriss² 2.3.619 ff., 624.

²⁴ In Tocharian B (or ‘Kuchean’) the material is scarce; no forms of either the 1st or 2d person are to be found in our texts, as far as I know. The 3d person forms correspond to those in A (considering that A drops the final vowels, which B usually preserves). As dialects A and B are on the whole very closely related, the paradigms of A may to a certain extent be considered as holding for B also. See the material in Lévi and Meillet, MSL 18.2 ff. (1914).

The brackets in the form weñās[ ] indicate that our ms. is cut at this point, and we do not know whether or not a -t followed the -s. We have no complete 2d. pl. form.

²⁵ In Slavic we also find a mixed paradigm, in which some forms have -s- and some do not, but the distribution is different (2d and 3d singular not sigmatic, all other persons sigmatic), cf. Leskien, Altbulg. Gramm.³ 201 f., 166.

On OE cnēow, sēow (perhaps = Lat. gnōuit, sēuit) see Leumann, Lat. Gramm. 335 (with bibliography); on blāwan (cf. flāuī?), Hirt, Idg. Gr. 4.268 f.; Handbuch des Urgerm. 2.148, 150 f.

Vendryes (Rev. celt. 44.258 [1927]) also comzgsb pares the mysterious Celtic ieuru (3d sg. perf.).

²⁶ The Latin 3d plural ending -runt is compared by Burger (REL 4.213 [1926]; Études 114) with Skt. -ran (cf. also Sommer, Handbuch³ 578); it could also, of course, be explained as a sigmatic form: *amā-s-ont, *-plē-s-ont, although this would present the disadvantage of separating it from the corresponding Hittite and Tocharian forms, which have -r-. The -unt can easily be explained, like the -an of Vedic -ran, as an imitation of the 3d plural present and future (and aorist).

²⁷ The of the 1st singular has been either preserved or rebuilt by the analogical influence of all the other Latin perfects in -ī. For early Latin, the type presents no phonetic difficulty; and for later Latin, its existence is proven by the testimony of Probus.

²⁸ There is, consequently, a radical difference on this point between Burger's idea and mine; for Burger considers -plēstī to be a Latin creation = 2d sg. -plēs -+- perfect ending -tī (!) as in dīxtī, lāuistī (Études 115; cf. also 105), whereas I consider the element -st- in the ending to be Indo-European, and compare it with the Hittite and Tocharian endings, which also contain -st-.

²⁹ See the material in Neue-Wagener 3.480. For compleō, we have always -plēsti(s) and -plēssent, never -plēuistis or -plēuissent (see TLL s.v. compleō). Likewise, I know examples of crēstī (cf. cernō), dēlēstī (cf. dēleō), (con-)quiēstī, (con-)suēstī (con-)suēstis, (ad-)suēstis, but none of *crēuistē, *dēlēuistī, *quiēuistī, *-suēuistī *-suēuistis (Plautus Cist. 87 can be read consuēstī, cf. Bauer 44, §36a).

³⁰ One of the Tocharian A preterite paradigms (III) has -w- in the 1st person, but not in the others; it also has -st- in the 2d singular (no examples of the 2d plural are attested; SSS 331, 375): thus yām-, 1st sg. yāmwā, 2d sg. yāmaṣt, 3d pl. yāmār, etc. In the middle we have 1st sg. yāmwe, 3d sg. yāmte; since e in Toch. A always represents an IE diphthong (ai, ei, oi), yāmwe exactly corresponds to a Latin perfect like -plēuī, sēuī, flēuī, etc., where -ī represents IE -ai, the ending of the 1st person perfect middle (Indo-Aryan cakr-é, OCS vědě, cf. OLat. inscr. fecei, petiei, posiuei 132 b.c., Faliscan peparai; see Sommer, Handbuch³ 574).

The ‘reduplicated’ Tocharian preterite (probably an old IE perfect) also has the element -w (-wā) in the 1st sg., but not in the others, cf. śaśmāwā(ṃ) (caus. of ṣtām-), etc., but 2d sg. śaśmāṣt, kakiāṣt (cf. käl-), śaśrāṣt (caus. of tsär), SSS 371; this type of preterite with reduplication can be compared even more directly with the Vedic perfects like papráu, jajñáu. In Tocharian B, no forms of the 1st person are known; see S. Lévi and A. Meillet, MSL 18.2 (1914).

In Armenian, inversely, -w is preserved in the 3d sg. but not in the 1st sg. (middle): cnay ‘I am born’, cnaw ‘he is born’; caneay ‘I have known’, caneaw ‘he has known’, etc. (see REA 10.183 f. [1930]; Esq. armén.² 124.).

I do not know at what Hittite facts Meillet is hinting in the last passage cited; probably at forms like the preterites kwenun, paun, uhun, tahun, spanthun etc. (1st sg.). IE *u is represented by u in Hittite (Sturtevant, Comp. Hitt. Gramm. 96) and IE w became u after a consonant at the end of a word; now, in all the Hittite forms except paun, -u is the ending of consonant stems. The -n of course represents the IE 1st singular ending *-m, and is a posterior addition. Cf. Benvéniste, Festschrift Hirt 2.230. (Other explanations, which do not convince me, may be found in Sturtevant, Comp. Hitt. Gramm. 254, with note 75.)

³¹ Burger (Études 121 ff.) also claims to find remainders of the old paradigm nōuī nōstī (with in the first and third singular only) in Ital. sęppi sapẹsti sęppe; tẹnni tenẹsti tẹnne; ębbi avẹsti ębbe (on the model of which were formed also dissi dicesti disse and similar ‘strong’ perfect paradigms). (To sapistī [= Ital. sapẹsti] and sapisset [= Ital. sapẹsse], add also resipistī [Plautus, Miles 345]; cf. Neue-Wagener 3.245 ff.) If Burger is right, the use of the perfect stem derived from the forms with in the Ital. 3d pl. tàcquero, , etc., about which he says nothing, is due to its accent, cf. , because of the opposition fēcī, fēcimus, fēcerunt: fēcistī, fēcistis (Ital. : facẹsti, facẹste; ębbi, ębbero: avẹsti, avẹste; Italian also has vẹddimo, , etc., which may be old, as ).

³² For other phenomena common to Latin and Indo-Aryan (and most of them to Iranian too), cf. Bonfante, I dialetti indo-europei 164 f. (Naples, 1931).

³³ The analogy of the paradigm nōuī nōstī nōuit can perhaps explain why the 1st and 3d sg. forms cantāuī cantāuit are much more frequent than the other forms with (cantāuistī cantāuimus cantāuistis ); they dominate almost without competition in our Latin texts, and cantāuit has perhaps even partly passed into Romance (cantáut, in Italian, Spanish, and Portuguese). (I owe this remark and a slight change in the text in part to a suggestion of Professor Sturtevant during the discussion of this paper at the meeting of the Linguistic Society in December, 1940.)

³⁴ In the Carmina latina epigraphica, the short forms of the 3d plural ‘rarissima sunt’ (Muller, Mnemosyne 56.370 [1928]); there are 6 cases of -ārunt, 1 of -ērunt, 1 of -ōrunt.

³⁵ Cf. Burger, Études 116: ‘Cette prédilection de la langue, fortement artificielle et pédante, de la chancellerie pour les formes longues ne prouve nullement qu'elles soient archaīques, mais seulement quelles passaient pour telles aux yeux des rédacteurs des inscriptions. Bien entendu, pour un Cicéron, nōsse était une forme abrégée (imminutum) de nōuisse (Orat. 47.157), comme meum factum était contracté (poeta qui . . . contraxerat . . .) de meōrum factōrum (ibid. 46.155)‘.

This is also confirmed by the great scarcity (6 occurrences) of -ērunt (instead of -ěrunt) in the Carmina epigraphica; see the material in Muller, Mnemosyne 56.369 (1928)—he seems to interpret the facts otherwise, see 373, lines 2 ff.; but it is perhaps an inexact expression; see also Bauer 70 ff., §§77 ff., whose view on this point I entirely share.

³⁶ Ancient Latin still distinguished the paradigm of the future perfect with from the perfect subjunctive with ī (cantārīs cantārīt cantārīmus cantārītis); see Sommer 583; Leumann 340.

³⁷ Cantāssō, leuāssō, negāssim, prohibēssīt, licēssit, ambīssint, mercāssitur, turbāssitur, impetrāssere, etc. (see the material in Sommer, Handbuch³ 585 ff.) are of course no more ‘contracted’ forms than are cantārō, cantārim etc., however they may be interpreted—as cantas-sō (aorist subjunctive; Brugmann, Gdr.² 2.3.390 f.), or through the analogy of amāssem or of capsō, faxō (cf. Sturtevant, CP 6.221 [1911], followed by Leumann, Lat. Gramm. 344 and by Bauer 44, fn. 11), or by ‘gémination expressive’ (Benvéniste, BSL 23.53; cf. also Burger, Études 130).

³⁸ In the redaction of this article I owe a great debt of gratitude to my friend Dr. Robert A. Hall Jr. of Brown University. It is a pleasure for me to thank him here most heartily.