¹ For convenience, I follow the practice of most scholars by including Hittite among the Indo-European languages and terming the Ursprache Indo-European rather than Indo-Hittite, although I am inclined to agree with Sturtevant that there is good reason not to do so. For the most recently published statements on this question, see the reprint of Sturtevant's earlier statement, Lg. 38.105–10, and Lane, Lg. 37.472, as well as Goetze apud Lane, ibid. fn. 12a. At all events the problem is irrelevant for the purposes of this paper.

² These articles, to which the reader of the present one is referred for fuller discussions, with more numerous examples and more complete documentation, are as follows. In TAPA : ‘Voice of non-finite verb forms in Latin and English’, 74.269–306 (1943) ; ‘Vestiges of partitive apposition in Latin syntax’, 84.92–123 (1953) ; ‘Partitive apposition in Homer and the Greek accusative’, 85.197–289 (1954). In Lg.: ‘Some Hittite-Sanskrit parallels: Part II. Double dative: Infinitive plus dative noun’, 29.246–51 (1953). In Proceedings of the Ninth International Congress of Linguists (Cambridge, Mass., 1962); ‘Partitive apposition’, 784–94. The last two articles are hereafter designated as Lg. 29 and Cong, respectively; references to volume number (74, 84, or 85), with or without following page number or numbers, are to the articles in TAPA. Other bibliographical data are to be interpreted as follows. AJP = American journal of philology. Bechtel = George Bechtel, Hittite verbs in -sk- (Ann Arbor, 1936). Benfey = Theodor Benfey, Handbuch der Sanskritsprache, 2 vols. (Leipzig, 1852–4). Brugmann = Karl Brugmann, Grundriss der vergleichenden Grammatik der indogermanischen Sprachen ² 1–2, 2 vols. (Strassburg, 1897–1916). Buck = Carl Darling Buck, Comparative grammar of Greek and Latin, 4th impression (Chicago, 1948). Buck, OU = Carl Darling Buck, A grammar of Oscan and Umbrian (Boston, 1904). Clark = M. Tulli Ciceronis orationes, Philippicae I–XIV, ed. by Albertus Curtis Clark (Oxford, no date, preface dated 1900). CP = Classical philology. Delbrück = B. Delbrück, Grundriss der vergleichenden Grammatik der indogermanischen Sprachen vols. 3–5, (Strassburg, 1893–1900). Ellis = Robinson Ellis, A commentary on Catullus ² (Oxford, 1889). Ernout and Thomas = Alfred Ernout and François Thomas, Syntaxe latine (Paris, 1951). Friedrich (or Friedrich, El.) = Johannes Friedrich, Hethitisches Elementarbuch, 1st part (Heidelberg, 1940). Friedrich, Vert. = Johannes Friedrich, Staatsverträge des Hatti-Reiches in hethitischer Sprache, 2 parts (Leipzig, 1926–30). Goodwin = William W. Goodwin, A Greek grammar, revised edition (Boston, 1895). Goetze, Hatt. = Albrecht Goetze, Hattusilis (Leipzig, 1925). Goetze, Madd. = Albrecht Goetze, Madduwattas (Leipzig, 1928). Goetze, NBr. = Albrecht Goetze, Neue Bruchstücke zum grossen Text des Hattusilis und den Paralleltexten (Leipzig, 1930). Grandgent = C. H. Grandgent, An introduction to Vulgar Latin (Boston, 1907). Hofmann = Manu Leumann and Joh. Bapt. Hofmann, Stolz-Schmalz Lateinische Grammatik ⁵ (Munich, 1928). JCS = Journal of cuneiform studies. Kent = Roland G. Kent, Old Persian (New Haven, 1950). Merrill = T. Lucreti Cari de rerum natura, ed. by William Augustus Merrill (New York, 1907). Morris = The Captives and Trinummus of Plautus, ed. by E. P. Morris (Boston, 1899). Mueller = M. Tulli Ciceronis scripta quae manserunt omnia, ed. by C. F. W. Mueller, 4 parts, 10 vols. (Leipzig, 1889–98). Munro = T. Lucreti Cari de rerum natura ⁴, ed. by H. A. J. Munro, 3 vols. (Cambridge, 1886). Ose = Fritz Ose, Supinum und Infinitiv im Hethitischen (Leipzig, 1944). Pedersen = Holger Pedersen, Hittitisch und die anderen indoeuropäischen Sprachen (Copenhagen, 1938). Peterson = M. Tulli Ciceronis orationes Verrinae, ed. by Gulielmus Peterson (Oxford, 1916). Prokosch = E. Prokosch, A comparative Germanie grammar (Philadelphia, 1939). Sommer, HAB = Ferdinand Sommer and Adam Falkenstein, Die hethitisch-akkadische Bilingue des Hattusili I (Labarna II) (Munich, 1938). Sommer, Heth. II = Ferdinand Sommer, Hethitisches II (Leipzig, 1922). Streitberg = Wilhelm Streitberg, Gotisches Elementarbuch ⁴ (Heidelberg, 1910). Studies Whatmough = Studies presented to Joshua Whatmough on his sixtieth birthday, ed. by Ernst Pulgram (The Hague, 1957). Sturtevant (or Sturtevant, HG ¹ or HG ²) = Edgar H. Sturtevant, A comparative grammar of the Hittite language: HG ¹ = first edition (Philadelphia, 1933); HG ² = revised edition (New Haven, 1951); all references to Sturtevant not otherwise designated are to HG ². Szantyr = J. B. Hofmann, Lateinische Syntax und Stilistik, revised by Anton Szantyr (Munich, 1965). TAPA = Transactions of the American Philological Association. Warmington = Remains of Old Latin, ed. and translated by E.H. Warmington, 4 vols. (Cambridge, 1935–40). Whitney = William Dwight Whitney, Sanskrit grammar ², 5th issue (Cambridge, 1923). ZA = Zeitschrift für Assyriologie und verwandte Gebiete. References are to pages. Quotations from illustrative passages are sufficient to give sense and syntax, but are not necessarily complete, and omissions are not indicated. In transliterations of Hittite (including Hittite names in titles of books), s is to be read as sn and h as hh.

³ Merely the present tense of Gothic, and a few other forms. See Prokosch 209.

⁴ See Grandgent 170.

⁵ In Germanic, mainly ‘get’ and ‘be’; in Romance, mainly ‘be’.

⁶ That there really is a difference between them, at least in Latin, is indicated by the fact that when the noun is replaced by a pronominal form, the subjective genitive is regularly represented by a possessive adjective, e.g. meus, noster, and the objective genitive by a genitive pronoun, e.g. mei, nostri. (Still a third variety of genitive, the partitive, is in the first and second person plurals represented by a different form of the genitive pronoun, e.g. nostrum instead of nostri.)

⁷ The rule is given particularly often with respect to verbal adjectives, i.e. participles: e.g. for Hittite by Friedrich 35 and 81 and Pedersen 148 (so too Sturtevant in HG ¹158, but, convinced by my arguments, he modified this statement in HG ² 78), and for Old Persian by Kent 78; similarly for Gothic by Streitberg 215. Much better statements are given concerning Sanskrit by Whitney 340 (cf. below, fn. 17), and concerning Sanskrit, Greek, and Latin by Buck 308, who says well of the Indo-European verbal adjective in -to- in general, ‘from transitive verbs it would usually have passive force, but it might also be formed from intransitive verbs’.

⁸ This distinction is of considerable significance and influence in some languages, such as Human and the North Caucasie group (cf. 74.270 and fn. 2); but in the Indo-European languages, in which often one and the same verb can be used both transitively and intransitively, it seems to me of no great weight. (For some variations in English, cf. fn. 11; for a possible effect of the variation in Hittite, cf. fn. 30.)

⁹ Cf. 74.272–4.

¹⁰ We do meet a few sporadic instances. I can cite the following: with a participle, Ann. 3.12 ^vdui-ya-an-za IGI.H.I.A-wa ^DUTU-i ne-an-za ‘a sheep turning (its) eyes to the sun’; with a gerund, KUB 2.1.2.26 ŠU-an ap-pa-an-na-as ^DKAL-ri ‘to the tutelary deity of grasping a hand’; with an infinitive, KBo. 4.4.2.63–4 ku-e KARAŠ.HI.A I-NA KUR ^URUNuhas-si hal-kiHIA-us har-ni-in-ku-wa-an-zi pí-e-hu-da-an har-ta ‘the troops which he had led into Nuhassi in order to destroy the crops’. Cf. 74.272–3.

¹¹ Much the same state of affairs still prevails to a great extent in modern English, millennia later. Our infinitives are frequently voiceless. Contrast the boys are ready to eat (eat intransitive, and seemingly active) with the apples are ready to eat (eat transitive, and seemingly passive) ; however, in the second instance we can substitute a true passive (periphrastic) form, and say the apples are ready to be eaten. Sometimes one and the same form is ambiguous, as the ship is ready to sail, where it is impossible to be sure whether the sense is that the ship is going to sail (intransitive, active) or to be sailed (transitive, passive); similarly, the apples are ready to roll. Gerunds are also voiceless, but I think in some instances we are less likely to use the voiceless form in an active sense, as the rather improbable the boys are ready for eating, than in a passive sense, as the apples are ready for eating. On the other hand, we have the voiceless form of the gerund in his hair needs cutting as opposed to the passive form of the infinitive in his hair needs to be cut. I discussed this and allied problems at some length over twenty years ago, 74.302–6.

¹² It also has a probably adjectival form in -wes resembling the Latin gerundive. This is dealt with below, fn. 90, where it is noted that this form too, like the participle (on which see fn. 7), seems active when it comes from an intransitive verb, and passive when it comes from a transitive one.

¹³ Yet even here we meet exceptions. Thus I can perceive no difference in voice between the supposedly active present participle in Vergil's volventibus annis (Aen. 1.234) and the supposedly passive gerundive in his volvendis mensibus only 35 lines later (1.269). Finite forms of verbs do not lose their voice so readily: if we substitute a finite verb for the participle, I think the verb would have to be the mediopassive, anni volvuntur, not the active, anni volvunt. See 74.276–7.

¹⁴ The active form is called present, and the passive form is called past or perfect, but the latter at least must have been originally tenseless as well as voiceless. Cf. fn. 15.

¹⁵ Both -to- and -no- were adjectival suffixes, which accounts for the fact that participles formed with them must have been originally both voiceless and tenseless. Rather surprisingly, both elements, while very productive in most Indo-European languages, seem to be lacking in Hittite (Sturtevant lists ‘a few traces’ of -to-, some ‘open to more or less doubt’, in HG ^l 157, and none at all in HG ²). In Sanskrit, the -na- forms are rare (see Whitney 340). In Germanic, the -to- formation is used for weak verbs, and the -no- formation for strong verbs (see Prokosch 205–6).

¹⁶ On the -to- form in particular, see 74.276 fn. 30.

¹⁷ This is the comment by Whitney which was referred to with approval above, fn. 7.

¹⁸ These examples are cited by Buck 308.

¹⁹ Verbs meaning ‘eat’ and ‘drink’ tend to be intransitive in Hittite also; cf. below, fn. 30. However, the Latin verbs to which the participles here cited belong can also be transitive, meaning respectively ‘dine upon, eat’, ‘breakfast or lunch upon’, ‘drink’ (with an object), ‘call upon by an oath’ or ‘bind by an oath’; with this meaning potus and iuratus are used in a passive sense also, and conceivably cenatus and pransus could be. Incidentally, English usage is parallel: cf. he has eaten (active) with the meal is eaten (passive).

²⁰ In Sanskrit the participles from these tense-stems are quite distinct from the voiceless -ta- and -na- participles, which are in general formed directly from the verbal root (Whitney 340).

²¹ Participles from all these tenses are met in Vedic Sanskrit, though later the perfect is rare (Whitney 292) and the aorist lacking (Whitney 202 and 220).

²² This, a formation in -māna-, is a true present and a true passive, and must not be confused with the so-called passive particple in -ta- and -na-; see Whitney 276.

²³ Both languages also, like Hittite (cf. fnn. 12 and 90), possess certain adjectival forms which are used much like the Latin gerundive, and which are sometimes classed as, or at least compared with, future passive participles (on this, however, see fn. 26). These include Sanskrit derivatives in -ya, -tavya, and -anīya (Whitney 345), and Greek verbals in -teos (Goodwin 167).

²⁴ Two paragraphs earlier.

²⁵ It is the only participle of the deponent verb that follows the regular rule for deponents : passive form, active meaning. The present and future participles of deponents are irregular in having active forms, and the gerundive in having passive meaning.

²⁶ Cf. fn. 23. However, I would certainly not classify the gerundive as a future passive participle, since its meaning usually involves necessity rather than simple futurity. Incidentally, there are instances of its use in an active sense, notably Plautus, Ep. 74 puppis pereunda est ‘the ship is (sure) to perish’ and Trin. 1159 si ilia tibi placet, placenda dos quoque “if she pleases you, her dowry too has to please (you)'.

²⁷ This is true also of Latin (cf. fn. 48). On the other hand, Greek, as we have seen, is rich in participles, but, as we shall see, poor in verbal nouns (cf. fn. 46). Sanskrit alone is rich in both (cf. fn. 42).

²⁸ See Sturtevant 68–77.

²⁹ See Sturtevant 71–5, 98 and 99.

³⁰ The two sets are certainly parallel in use. On the genitives in -was and -annas, see fn. 90; on the datives in -wanzi and -anna, see fn. 35, also Ose passim, especially his conclusion (87–8), where he also touches on the respective origins of the rival forms, suggesting that originally the -anna form belonged mainly to the -mi- class of verbs. Goetze makes the distinction on quite different grounds: he considers the -wanzi form, and indeed all the -war forms, active, and the -anna form, and indeed all the forms in -atar, mediopassive (cf. NBr. 31, and the paradigms in Madd. 93–5). Under his influence, Sturtevant in HG ¹ 267 stated that as a rule the forms in -war and -wanzi belong to active verbs, and those in -tar and -anna to middle verbs. Goetze in JCS 2.151–5 reaffirmed his earlier view, in somewhat modified form; but Sturtevant in HG ² 73 specifically abandoned his, saying that some forms in -tar -nnas are connected with active verbs (which of course Goetze realized too, loc.cit.). Sturtevant's other reason for reversing himself was his acceptance of my view that Hittite verbal nouns lack voice, but that is not a cogent argument in this instance; the two forms might, as I believe, mean exactly the same thing, and yet one of them might be associated primarily with the ‘active’ conjugation and the other with the ‘mediopassive’ (as indeed can be true of finite forms also). I have ventured to wonder (despite my observations in fn. 8) whether the distinction may be that, though there are certainly exceptions (e.g. from ap- the genitive appannas, as in KUB 2.1.2.26, quoted in fn. 10, and the dative appanna, as in the forms cited by Goetze, JCS 2.153), Hittite tends to use -wanzi from transitive verbs and -anna (as well as other forms of -atar) from intransitive verbs. This would account for two types of -atar particularly mentioned by Goetze in JCS 2.153–4 and 154–5, namely: (a) from ak- ‘die’ and its synonym hark- ‘perish, disappear’, (to which I would add hwes- 'live ; cf. TI-annas in Hatt. 1.15, discussed below, fn. 90); (b) from ed- 'eat', wak- ‘bite, have a bite’, aku- ‘drink’, all of which Goetze (loc.cit.) specifically states are not used with accusatives. My suggestion of course does not mean that the -wanzi form itself is regularly used with an object; actually, this happens very rarely (the instance quoted above in fn. 10, KBo 4.4.2.63–4, is, according to Goetze, JCS 2.146, ‘the best example‘). As for the -anna form, Goetze in opposition to Ose reiterates (JCS 2.151) a view that he had expressed eighteen years earlier (in NBr. 31), that it is never used with an object. Friedrich (80) declares that neither of the two can take an object, but that is certainly not true of the -wanzi form.

³¹ Sturtevant in HG ¹ (154) suggested that these forms are datives of a -t- extension of the ~r/n- stem seen in -war, and this still seems plausible to me, though he did not repeat it in HG ², saying there (75) merely that the -i follows a -t- extension, without explaining the preceding -wan. The -zi he explains elsewhere (60) as the regular development of -ti; in fully declined -nt- stems the retention (or restoration) of the original -ti of the dative singular is probably due to the influence of the other case-forms, but Sturtevant holds that the isolation of the infinitive accounts for its -zi (75). Pedersen, however, queries (150) whether the -wanzi form may be a contamination of -wan and a vanished -zi; in my opinion this, as he himself realizes, is highly dubious.

³² Sturtevant (74) calls it ‘apparently a suffixless dative or locative in origin’, and Pedersen (150) categorically pronounces it a suffixless dative. However, I can see no objection to taking it as the accusative of the -nt- extension of the noun which Sturtevant once posited as the starting-point of the infinitive in -wanzi (see fn. 31). Syntactically, it certainly behaves like an accusative (see fn. 78).

³³ See Sturtevant 73 and 74.

³⁴ An added advantage is the parallel with Latin terminology, since probably these verbal nouns generated a gerundive just as I believe the Latin gerund did; see above, fn. 12, and below, fnn. 87 and 90. It is for this reason that Goetze (Hatt. 140) compares the genitive form in -was with the gerund, although he calls the nominative form in -war (as well as that in -atar) an infinitive (Madd. 93–4, reaffirmed JCS 2.155). Friedrich too calls the -war form an infinitive (36, 64, and 80) ; the -was form he regularly terms simply the genitive of the infinitive (36 and 64), but he also, like Goetze, refers to it as a gerund (64). Sturtevant rejects the term infinitive for the -war form on the ground that verbal noun describes its use more accurately (74), and he employs this term for both the -war and the -atar forms (148), and calls the -was form the genitive of the verbal noun (74, 148). Similarly, Pedersen calls the -war and -atar forms verbal substantives (149), and the -was and -annas forms the genitives (149) and the -anna form the dative (150) of these verbal substantives. Sommer also terms the -war form a verbal substantive (HAB 226). However, I do not consider the choice of these scholars a happy one, for the forms that they all call infinitives and supines are verbal nouns just as the -war and -atar forms are.

³⁵ Goetze stated long ago (NBr. 31) that the forms are syntactically equivalent. Friedrich (80) said of their parallel use that the precise conditions were ‘noch zu untersuchen’. Such an investigation was subsequently carried out by Ose in his excellent monograph, and his research certainly justifies his conclusion (87) that the two are in complete agreement and can be employed in ways ‘ganz parallel’.

³⁶ Sturtevant uses the term infinitive for both the -wanzi form and the -anna form (148), distinguishing them in his paradigms (e.g. 151, 152) as infinitive a and b respectively. So too Ose, who has proved so cogently their complete parallelism (cf. fn. 35 above), would like to see them both called infinitives (88 fn. 8). Goetze in his review of Ose (JCS 2.155) agrees that the same term should be used for both, but he prefers to reserve the term infinitive for the forms in -war and -atar (as noted in fn. 34), and to apply to the forms in -wanzi and -anna the term supine, which Ose, 88 fn. 8, advocated confining to the form in -wan. While I agree with Goetze that the choice of terminology does not make much difference, it occurs to me that applying the same term to the two synonyms, and also to the very differently used form in -wan, might lead to inconvenience.

³⁷ While, as I have indicated, I can see advantages in associating the -anna form by name either with the -atar form, to which it is related formally, or with the -wanzi form, to which it is parallel syntactically, I cannot approve a system of nomenclature that groups either -anna or -wanzi specifically with any other verbal noun. Thus I disagree with Friedrich (36, 80), who calls the -war form and the -anna form both infinitives, first and second respectively, and who also calls the -wanzi form and the -wan form both supines, first and second respectively. Sommer, too, classes both the -wanzi and the -wan forms as supines, but, reversing Friedrich, calls them second and first supine respectively (HAB 226).

³⁸ In support of this, it may be noted that in Latin, the infinitive and case forms of the gerund are not infrequently used as parallels or equivalents. See 74.287 fn. 89; also Szantyr 348 and 378–9. In English too the infinitive and the gerund are often to some extent, though not completely, interchangeable; e.g. to see is to believe and seeing is believing, I like to walk and I like walking. On some differences between them, cf. fn. 11.

³⁹ Sturtevant 75 and 149; Friedrich 36, 63, and 80; Pedersen 150; Sommer, HAB 226.

⁴⁰ See the examples listed in fn. 10. The supine in -wan also seems to take an object, but here another explanation is possible and, I think, preferable, as I shall show below; see fnn. 78 and 79, and the portion of the text to which they refer.

⁴¹ See Whitney 203 and 347–60.

⁴² Sanskrit is the only one of the four languages here studied that is rich in both verbal nouns and verbal adjectives. See fn. 27.

⁴³ The ablative and genitive infinitives are much restricted in their use. See Whitney 353–4.

⁴⁴ See Whitney 355–60.

⁴⁵ See Whitney 351–5, 358–60.

⁴⁶ Cf. fn. 27.

⁴⁷ Brugmann 2.3.919–21 does group according to case a number of infinitives without articles, but his classification, while probable, is not really provable. Cf. fn. 69 below.

⁴⁸ Cf. fn. 27.

⁴⁹ The nominative gerund is largely confined to the early period (cf. fn. 51, and see 74.286–7 and 297–8). Most grammarians deny its existence. For instance, the late Joshua Whatmough, in one of his last publications (CP 59.112), stated categorically, 'The gerund has no nominative.' I wish I could have asked him how he explained, e.g., Plautus, Trin. 869 agitandumst vigilias. I discuss this question in detail in a paper, 'Was there a nominative gerund?', to appear in TAPA 96.

⁵⁰ The dative supine is not universally recognized; but see 74.298–9 and 84.117 fn. 123, and cf. fn. 76 below.

⁵¹ The Latin infinitives were datives and locatives in origin, but, lacking the flexibility which the existence of the definite article gave to Greek, they came to be almost exclusively confined to nominative-accusative uses. In other cases, the place of the infinitive was largely taken by the gerund, which accordingly lost through desuetude its nominative and, except with prepositions, with which a Latin infinitive was extremely rare, its accusative. Cf. fn. 49.

⁵² In origin the Latin infinitive must have been voiceless like other verbal nouns. The difference between the present active and the present passive infinitives was originally one of case, not voice: the active form in -re (earlier -se) was once a locative, and the passive form in -ri (earlier -si) or -i was once a dative. Cf. Buck 305–6.

⁵³ We generally think of the Latin gerund as active in opposition to the gerundive, but actually a word that does not vary in form to denote variation in voice can hardly be said to possess voice. Just as the gerundive is not always ‘passive’ in sense (cf. fn. 26), so the gerund is not always ‘active’. For instance in Vergil, Georg. 3.215–6 urit videndo femina: videndo certainly has passive implications; the female sets the male ablaze not by seeing him (videt femina) but by being seen by him (videtur femina). Servius on Vergil, Ecl. 8.71 aptly contrasts cantando in Ecl. 3.25 cantando tu ilium! (an incredulously ironical aposiopesis for num cantando tu illum vicisti?) with the same word in 8.71 cantando rumpitur anguis (the snake is caused to burst not by its own singing, but by being sung at, i.e., bewitched; for canto used transitively in this sense, cf. Propertius 4.5.13 cantatae lunae). For further examples, see 74.277 and fn. 36. It should be noted that in all the three Vergilian passages just quoted, the gerund is to be interpreted in terms of the subject of the clause, but sometimes this is impossible, as in Plautus, Men. 882 lumbi sedendo dolent, and Terence, Phorm. 1034 accusando fieri infectum non potest (cf. the French l'appétit vient en mangeant, frowned upon or apologized for by purists); here the agent involved is not mentioned at all, and I suppose we would have to call the gerunds ‘active’, unless we wish to say that they correspond to the impersonal ‘passive’ sedetur (used as is the familiar pugnatur) and accusatur (in French the corresponding finite form would be the active impersonal on mange). And of course we could treat cantando in Ecl. 8.71 in the same way. (Incidentally, once more when the verb involved is intransitive, the nonfinite form seems active; and when the verb is transitive, it seems passive. But actually cantando and accusando have no more voice than cantu or accusatione.)

As for the voice of Latin supines, this too is frequently a matter of analysis or interpretation. If we associate the supine closely with the noun in the same case, we find that with a dative (intransitive) supine, this noun is in subjective relation to it, and with an accusative (transitive) in objective relation; so the intransitive supine seems active, and the transitive supine seems passive. Thus in Mil. 724 amicis usui est ‘the friends use him’; but in Bacch. 347 amicos iit salutatum ‘the friends were saluted by him’. (These two passages are cited again in Part III as examples in Latin of the dative and accusative respectively among instances recalling the original use of a ‘whole-noun’ and a verbal ‘part-noun’ in partitive apposition.) But if we associate the supine rather with the subject of the clause, the situation is reversed : he is used by his friends, he went to salute his friends. An interesting example is Cato, Agr. 5.3 satui semen, cibaria, far, vinum, oleum mutuum dederit nemini ‘let him give seed for sowing, fodder, spelt, wine, (or) oil as a loan to nobody’. Here the occurrence after semen of a number of other accusatives to which satui could not apply forces our associating the supine with the nominative semen, and thus satui seems passive; but if we had had only satui semen dederit nemini, with no other nouns interpolated, we might have associated the supine with the other dative nemini, in the sense ‘let him give grain to nobody for sowing’, and thus satui would have seemed active; but really it is voiceless, just as the English sowing is. Probably in most instances of the use of a ‘whole-noun’ and a verbal ‘part-noun’ in the same case, we might find it possible to vary our interpretation of the ‘voice’ of the verbal noun as the result of a shift in analysis of this sort—which is one more proof of its actual voicelessness.

⁵⁴ In this respect they are certainly unique among ancient verbal nouns. In modern English, I think an infinitive can likewise take a subject. Thus in such a sentence as I believe the man to be mistaken, I would call man the subject of the infinitive (it is certainly not the object of the main verb), although the fact that we cannot say the man to be mistaken is believed perhaps constitutes an objection to this explanation.

⁵⁵ The development of the infinitive with subject accusative in indirect discourse presents an interesting problem. On this see fn. 104.

⁵⁶ It is doubtless for this reason that the accusative supine is considered active and the others passive. Cf. fn. 53.

⁵⁷ There was certainly no objection in the early period to the gerund with an object (see 74.295–8). I cite just one example for each case, all from Plautus: nominative, Trin. 869 agitandumst vigilias (already quoted in fn. 49); genitive, Cas. 190 ius optinendi optio; dative, Ep. 605 Epidicum operam quaerendo dabo; accusative, Poen. 599 ad hanc rem agundum (some editors amend to agundam, but this seems unnecessary, and I think the lectio difficilior is to be preferred); ablative, Men. 883 manendo medicum. Even in the classical period, examples are found; see fn. 94.

⁵⁸ For a possible reason for this preference, see 74.297–8.

⁵⁹ See particularly 85 (on Greek) and 84 (on Latin).

⁶⁰ Such genitives are commonly classified in various uses, notably as subjective, objective, and partitive. Cf. fnn. 6 and 61.

⁶¹ Of course we have instances of this when the part-noun is a verbal noun. In such instances the whole-noun, if it performs the action, is in the relation to the part-noun of a subject to a verb, and the genitive by which it is ultimately replaced is ‘subjective’; if it receives the action, the whole-noun is in the relation to the part-noun of an object to a verb, and the genitive by which it is ultimately replaced is ‘objective’. For convenience, I term the two types of whole-nouns also ‘subjective’ and ‘objective’.

⁶² Notably in the largely archaic Law Code, the two may alternate in parallel passages, or even in variant versions of the same passage, with not the slightest distinction in meaning that I can detect. Thus we can talk about injuring in some way ‘a man his ear’, as in §16, or ‘a man's his ear’, as in §15 (the possessive ‘his’, which from our point of view is redundant, is extremely common in such instances, though not completely obligatory). For further details on partitive apposition in Hittite, see 84.95 with fnn. 12, 13, and 14; and for some additional examples, Lg. 29.249 and Cong. 785–6 and 791–2.

⁶³ An examination of Homeric usage makes this perfectly plain. I choose just one illustration to employ here. If a Greek wished to talk e.g. about ‘striking a man's belly’ or ‘pleasing a man's heart’, several alternatives were open to him. Either ‘man’ or ‘belly’ might become the complement of ‘strike’ (which governs the accusative), and either ‘man’ or ‘heart’ might become the complement of ‘please’ (which governs the dative). The other member of the phrase would then be some sort of modifier. With ‘striking a man’ or ‘pleasing a man’, ‘belly’ or ‘heart’ might be replaced by an expression of specification, either dative or accusative, the former an inheritance from the construction of the Ursprache, the instrumental, and the latter an essentially Greek innovation, appropriately known as the ‘accusative graecus’. With ‘striking a belly’ or ‘pleasing a heart’, ‘man’ might be represented by a dative of reference (common with body-parts in modern French and German also, whereas English uses the genitive). But I have noted a marked tendency on Homer's part to use either two accusatives or two datives, suggesting that for him the construction is still partitive apposition. Thus I think Il. 13.506 Oinómaon bále gastéra means not ‘struck Oenomaus in respect to (his) belly’ but ‘struck Oenomaus (his) belly’, and Il. 1.24 Agamémnoni hendane thumôi means not ‘pleased Agamemnon in respect to (his) heart’ or ‘pleased the heart with reference to Agamemnon’ (anglice 'pleased Agamemnon's heart'), but 'pleased Agamemnon (his) heart'. See 85.219 fn. 67 and the cross references there given.

⁶⁴ I have grouped some parallel forms from Hittite, Greek, and Latin in Cong. 786–7. For longer lists of Greek examples, see 85: genitive, 212–4; dative, 214–9; accusative, 219–39. For longer lists of Latin examples, see 84: dative, 105–7 (for the so-called double dative, 107–11; on this cf. inf., fn. 72); accusative, 101–3. Latin examples involving ablatives are rare (cf. 84.96), and apparently there are none involving genitives (cf. 84.96 and fn. 20).

⁶⁵ The second is far the commoner type; and in the instances that follow, it may always be assumed to be exemplified unless there is a special comment to the contrary (fnn. 71, 77, 81).

⁶⁶ On the special difficulty presented by Greek, see fn. 47.

⁶⁷ For a longer list and a more detailed discussion, see Cong. 791–4.

⁶⁸ Already quoted in fn. 10.

⁶⁹ This is one of a group of examples listed by Brugmann, 2.3.920, to exemplify this construction. I believe he is probably right in calling bouleûsai a genitive, but I cannot prove it, and neither can he. See fn. 47.

⁷⁰ Cf. above, for the name fn. 36, for the form fn. 31.

⁷¹ On the subjective type, cf. fn. 65.

⁷² In this connection it may be of interest to note that in Latin, of all survivors of partitive apposition in general (not necessarily those involving a verbal noun) those in the dative are by far the commonest; indeed, they occur so often that, though not generally recognized as remains of partitive apposition, they have been recognized as a special construction and given a special name, the ‘double dative’ (see 84.107–11, Lg. 29.248–9).

⁷³ See e.g. Brugmann 2.3.917–8, Delbrück 4.470, Whitney 352. That in such passages as RV 1.24.8 the whole-noun would have to be replaced not by the object but by the subject of the infinitive (cf. fn. 71), a construction impossible in Sanskrit, is ignored by most authorities; Whitney, however, with characteristic keenness, takes cognizance of it (loc.cit.). Both Brugmann (2.3.918) and Delbrück (4.471) accept Benfey's view (1.432) that the Vedic construction resembles the Latin gerundive, but what it really resembles is the Latin gerund as I believe we meet it in As. 250 (see fn. 75), or, in the comparatively rare instances where the whole-noun is subjective, the Latin supine as we meet it in Mil. 724 (see fn. 77). I discuss the Vedic construction at some length in Lg. 29.246–51; see also 74.274 fn. 25, and 281 with fn. 56.

⁷⁴ This is another of Brugmann's examples (2.3.920) ; cf. above, fn. 69. Of course here we might view the infinitive as simply used to express purpose, with no hint of original partitive apposition; but I think the parallelism with the examples from other languages here cited may justify Brugmann.

⁷⁵ I am assuming that such an expression as comparando here was originally a dative gerund used exactly as the genitive gerund tuendi is in Capt. 1008. Its eventual metamorphosis into a gerundive will be taken up later; see especially fn. 87.

⁷⁶ On the dative supine, see fn. 50.

⁷⁷ On the subjective type, again cf. fn. 65; also, for a Sanskrit dative of this type (RV 1.24.8), fnn. 71 and 73.

⁷⁸ See fn. 32. Its frequent combination with the accusative of the ‘whole-noun’, and its similarity in use to the accusative -tu- supine of Sanskrit and its Latin cognate the accusative supine in -tum, both support my view.

⁷⁹ As in Hatt. 4.4, cited above as an example of the dative use.

⁸⁰ Friedrich (63) accepts all three, but most scholars accept only one or two. See Bechtel 67–9, and the references there given.

⁸¹ On the subjective type, cf. once more fn. 65.

⁸² I realize that by Plautus' day amicos had undoubtedly come to be felt as the object of salutatum, but I trust that my explanation of its possible origin as an accusative of limit of motion will not be thought too daring. Both Sanskrit and early Greek freely use words referring to persons as accusatives with verbs of motion (see for Sanskrit Delbrück 3.363 and Whitney 92, and for Greek Delbrück 3.364; for the latter I cite as examples Homer, Il. 1.139 and Od. 1.332). In Latin the occasional use in this way of the name of a people is generally explained as derived from such a use of the name of a country, e.g. Vergil, Ecl. 1.64 ibimus Afros and 65–6 venimus Britannos, where it is doubtless significant that Britannos is parallel to Scythiam and even to the river name Oaxen (in Studies Whatmough 53–64, I have discussed Vergil's habit of confusing geographical and ethnological terms). But I think more weight may be attached to a passage in Gellius, 2.29.7 fac amicos eas et roges, where it is perhaps easier to take amicos directly with eas than to assign it to roges. What is important here is that Gellius is retelling a fable of Aesop which Ennius had retold before him: cf. 2.29.20 hunc Aesopi Apologum Q. Ennius in Satiris versibus quadratis composuit. In regard to this Warmington says (1.389) : ‘There can be no doubt, on the evidence of words and phrases of archaic look and of trochaic metre in Gellius’ reproduction, that Vahlen is right in believing that Gellius' prose reproduces Ennius' verse though it hides the metre almost completely.' If I am right that verbal nouns did not originally take objects, we might say that the Latin accusative supine borrowed its direct object from the accusative gerund and even from the infinitive (with which it is sometimes interchangeable) ; but it seems much more likely to me that the use of an accusative supine with an accusative noun was inherited, in view of the close parallelism of the Latin supine in -tum both in form and in syntax with the Sanskrit accusative infinitive in -tu-, and above all with the Umbrian accusative supine in -to, which occurs five times in the Tab. Ig. (lb 10, 6a 1, 5–6, 6–7, 6b 48), always depending on the verb 'go' and always accompanied by an accusative noun. It should also be noted that in early Latin not only the accusative supine is common with verbs of motion, but so too are other accusative nouns without a preposition, notably with eo, especially exsequias and infitias, sometimes too malam crucem and malam rem.

⁸³ I believe it was originally a gerund (cf. fn. 75), though by Plautus' time it may have already developed into a gerundive. On the accusative gerund without a preposition, cf. fn. 51, and see 74.291–3.

⁸⁴ Cf. fn. 69.

⁸⁵ Partitive apposition also, in my opinion, left vestiges of various kinds not involving verbal nouns and adjectives, the most striking being the accusativus graecus. These are treated in 84 and 85, also in Cong., but do not concern us here.

⁸⁶ Just as I believe the Greek accusative was (see 85.254–89, especially 282–6, and cf. fn. 85 above), and also the Latin infinitive in indirect discourse (cf. fn. 104 below).

⁸⁷ Namely, the shift from gerund to gerundive. This is treated in considerable detail in 74.279–94 and 84.114–7.

⁸⁸ Cf. 74.282–4.

⁸⁹ The entire fragment runs as follows: ita et fletu et tenebrie obstinatus speciem amisi luminis conspiciendi insolentia. Warmington translates this as follows (2.415): 'Held steadfast thus in grief and dark, I lost/ The light of eyesight through disuse of vision.' Evidently he took luminis with speciem, but to me it seems much more effective if combined with insolentia; speciem luminis would I think mean literally 'the sight of the eye', and would be practically redundant. Species alone means ‘sight’, as in Lucretius 4.236 and 242, 5.707 and 724; but luminis insolentia ‘unaccustomedness of light’ provides an effective reinforcement of tenebris obstinatus.

⁹⁰ In much the same way the Hittite gerund (see fn. 34) evidently generated a gerundive. To trace this development we begin with the ambiguous form in -was, seen in KBo 5.9.3.7 (the Duppi-Tesupas Treaty) me-mi-ya-as-ma ku-is i-ya-u-wa-as ‘whatever thing (is) of doing’ or KUB 1.1.4.55–6 (Hatt.) na-ah-hu-u-wa-as-ma-mu ku-is LUGAL-us e-es-ta ‘whatever king was of showing respect to me’. This form must have started as the genitive of a verbal noun (gerund), but in the passages quoted, it could be interpreted as the nominative of a verbal adjective (gerundive), so that the meaning would be ‘whatever thing (is) to be done’ and ‘whatever king was (supposed) to show respect to me’. As long ago as 1922, Sommer (Heth. 2.41 fn. 1) very tentatively suggested that iyawas might be a verbal adjective, and queried whether it was connected with the forms in -wan and -wanzi. Friedrich (ZA 1.10 and Vert. 1.45 and 2.187) maintained it was a noun, genitive of the infinitive as he called it (cf. fn. 34 above). Goetze, though he followed Friedrich in calling nahhuwas in Hatt. 4.55 the genitive of an infinitive in his commentary (Hatt. 102) and vocabulary (126), in an addendum on the passage (140) pronounced iyawas (and therefore presumably nahhuwas) nominative, i.e. an adjective (gerundive). In a sense, both are right. Friedrich (ZA 1.9) cites some passages in which the -was form cannot possibly be an adjective, as it does not agree with the noun which it modifies: an accusative in KBo 6.3.3.48 (Law Code §64) ták-ku ANŠU.KUR.RA tu-u-ri-ya-u-wa-as ku-is-ki da-a-i-iz-zi ‘if any one steals a horse of [i.e. for] harnessing’; a neuter in KBo 5.1.4.15–6 ar-ru-um-ma-as wa-a-tar ‘water of [i.e. for] washing’. He is right, of course, although not, I think, in his extra objection in regard to the second passage that ‘zu waschendes Wasser’ makes no sense (he fails to realize that ar-ru-um-ma-as, whether a noun or an adjective, can be active as well as passive). But Goetze (Hatt. 140) cites a passage in which the form under consideration cannot possibly be a noun. Beside the ambiguous form in KBo 2.2.3.33 ku-is IK-RI-BU sar-ni-in-ku-wa-as ‘whatever votive offering is of giving as atonement’ or ‘whatever votive offering is to be given as atonement’, we have the plural (and therefore obviously adjectival) form ibid. 4.7–8 IK-RI-BI.HI.A-ma ku-i-e-es sar-ni-in-ku-e-es ‘whatever votive offerings (are) to be given as atonement’. Friedrich in ZA 1 had dismissed the evidence of sarninkwes with the dictum (10 fn. 1), ‘Hoffentlich findet diese isolierte Form bald ihre Aufklärung’, regardless of its perfectly plausible ‘Aufklärung’ as a plural adjective; but he admitted in Vert. 1.45 that Goetze might be right about this form, and finally in El. 64 fully accepted Goetze's view as to the whole construction. So too Pedersen (149) and Sturtevant (74), although the latter makes the alternative suggestion (far less plausible to my way of thinking) that sarninkwes may be an old genitive singular. Finally, I should like to call attention to the fact that for iyawas and nahhuwas, from the gerund in -war, we can find parallels from the gerund in -atar. Goetze (JCS 2.152) compares two different gerunds from the same verb, KUB 26.58.1.10 ANŠU.KUR.RA wehuwas and KBo 6.28.2.25 ANŠU.KUR.RA wehannas. Since I can cite no instance of a form from -annas in a case or number demonstrably other than genitive singular, I cannot prove that this gerund too generated a gerundive ; but certainly TI-annas in Hatt. 1.15 Ú-UL-wa-ra-as TI-an ra-as seems just as much an adjective as TI-anza in 1.17 nu-wa-ra-as TI-an-za ‘he (will be) living’, and so the passage may mean ‘he (is) not destined to live’ (gerundive) rather than ‘he (is) not of living’ (gerund). Similarly KBo 4.10.1.10 ma-a-na-as har-ka-an-na-as na-as har-ak-du may mean ‘if he is (supposed) to die, let him die’ rather than ‘if he is of dying, let him die’. Note that, naturally, these gerundives are voiceless just like the gerunds from which they are formed. Thus iyawas and sarninkwas seem passive because they are from transitive verbs; nahhuwas, TI-annas, and harkannas seem active because they are from intransitive verbs (-mu with nahhuwas must be dative, not accusative).

⁹¹ Gerundives are regularly passive. However, active ones do occur in early Latin; two were cited above in fn. 26.

⁹² The reason for attributing priority to the gerundive is its existence in Oscan and Umbrian (Buck, OU 181); but the remains of these two languages are so scanty that the nonsurvival of the gerund need not give us pause.

⁹³ These include: (1) forms with the genitive of the personal pronouns as whole-‘nouns’, which, ending with -i, looked as if the part-nouns agreed with them, as Ovid, Her. 20.74 copia placandi tui (feminine singular), and Caesar, Bel. Gal. 7.43.2 sui purgandi gratia (masculine plural) ; (2) forms with the genitive plural of first and second declension nouns as whole-nouns, with which repetition of the forms in -orum and -arum was avoided, as Cicero, Inv. 2.2.5 exemplorum eligendi potestas, and Lucretius 5.1225 poenarum solvendi tempus. See 74.282–3 and 283 fn. 70.

⁹⁴ These include mainly: (1) forms with the neuter adjective or pronoun as whole-‘noun’, which if shifted in case from the accusative to fit the gerundive construction would not reveal gender, as Cicero, Inv. 1.25.36 aliquid faciendi ratio (the citation of this as consilium aliquid faciendi by Hofmann 596 and Szantyr 372 is an error), where if alicuius faciendi ratio were used instead, alicuius would seem like a masculine, and the result would be senseless; (2) forms with the genitive plural (like the second set cited in fn. 93), as Caesar, Bel. Gal. 4.14.2 arma capiendi spatio (paired with a gerundive consili habendi), Gellius 5.10.5 causarum orandi cupiens (paired with a gerundive eloquentiae discendae). In Cicero, De or. 2.38.157 artem vera ac falsa diiudicandi there is a double reason for the use of the gerund : the gerundive construction, artem verorum ac falsorum diiudicandorum, would have been both ambiguous and cacophonous. Varro, who frequently archaizes, seems fond of the gerund construction with prepositions, as in LL 9.42 ad discernendum figuras, RR 1.23.6 alia ad serendum.

⁹⁵ To be sure, the gerund construction populum Romanum opprimendi in this passage has been questioned. The whole passage, as recorded in the codices, runs as follows : agitur utrum M. Antonio facultas detur opprimendae rei publicae, caedis faciendae bonorum, urbis, agrorum suis latronibus condonandi, populum Romanum Servitute opprimendi, an horum ei facere nihil liceat. Manutius shifted to the gerundive construction, changing populum Romanum to populi Romani, and a number of modern editors, though by no means all of them, have accepted the emendation; for instance, Mueller, editor of the Teubner text, pronounces the conjecture ‘certissima’ (2.3.112), whereas Clark, editor of the Oxford text, preserves the reading of the MSS. It is true that no other example of the accusative gerund with a direct object is citable from Cicero, but direct objects with other cases of the gerund do occur in his works; and I for one am willing to accept the recorded lection.

⁹⁶ Cf. 84.118–9.

⁹⁷ A passive verbal adjective could have referred only to me, which of course was not in agreement with the verbal noun (cf. the passive participle with me in Plautus, Cas. 189 vir me habet despicatam). In the same way in Cato, Agr. 5.3 satui semen (already quoted in fn. 53), satui is practically equivalent to a passive verbal adjective (in this instance the gerundive serendum) in agreement with the subject semen.

⁹⁸ See 84.120–1 and Cong. 795 fn. 54.

⁹⁹ On this interpretation of the original construction of ancillulam here, and of gladiatores in Hec. 40 (quoted just below), see fn. 82 on that of amicos in Bacch. 347.

¹⁰⁰ On gladiatores see fn. 99.

¹⁰¹ A periphrasis of quite a different type was usually substituted for it: rumor venu fore ut dentur gladiatores.

¹⁰² If I am right as to its case. Cf. fn. 32.

¹⁰³ I have expressed my objection to the theory at greater length in my review of Ernout and Thomas, Lg. 30.259. See also 84.122 and Lg. 29.251.

¹⁰⁴ This construction, according to the usual view, is supposed to have generated the indirect discourse construction dico eum venire (see e.g. Hofmann 583, followed with slight modification by Szantyr 353) ; but this explanation I consider even less convincing than that of Ernout and Thomas, for if it were correct dico eum venire would have to mean not ‘I say that he is coming’ but ‘I tell him to come’. My own belief is that the oratio obliqua construction started not with an infinitive but with a past passive participle, subsequently misunderstood; see my discussion in TAPA 81.117–29. I was happy to have this followed up by a new proposal for the development of the similar but not identical construction in Greek indirect discourse by A. C. Moorhouse in AJP 76.176–83. I was also gratified to see an expression of agreement with my view by James W. Poultney in AJP 86.214.

¹⁰⁵ Even though it did, as I have said (fn. 51), come to function in the sentence usually as a nominative or an accusative rather than as a dative.