on the origins of categories of voice and aspect

7/29/2019 On the Origins of Categories of Voice and Aspect

1/14

Linguistic Society of America

On the Origin of the Categories of Voice and AspectAuthor(s): H. V. VeltenReviewed work(s):Source: Language, Vol. 7, No. 4 (Dec., 1931), pp. 229-241Published by: Linguistic Society of AmericaStable URL: http://www.jstor.org/stable/409225 .

Accessed: 08/10/2012 09:49

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access toLanguage.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=lsahttp://www.jstor.org/stable/409225?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/409225?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=lsa


2/14

ON THE ORIGIN OF THE CATEGORIESOF VOICE ANDASPECTH. V. VELTEN

STATE COLLEGE OF WASHINGTON[Cf. the opening and the concluding paragraphs.]

In two recent studies I have attempted to show (a) how such ubi-quitous grammatical categories as gender, number, case, person, andthe parts of speech are interrelated and have given rise to new cate-gories by mutual superimposition1; (b) that the aspects of the Indo-European verb go back to three older verb classes, namely (I) verbswhich express a physical or mental state like sleep, thrive, hate, (II)verbs designating a transition from one state into another: awaken, die,and (III) verbs indicating an action proper, i.e., an action caused by thevolition of the subject: run, seize, give2. It is the purpose of this paperto form a connecting link between these studies by showing that thethree original IE verb classes are inextricably bound up with the cate-gory of voice or diathesis, which in turn has its origin in the distinctionof the nominal and verbal phrase.In many African and American languages3 a 'verbal phrase' can berecognized as such grammatically only if it contains an object. Theobject can be expressed by a double pronominal affix of the verb as inNahuatl ti-nets-itta 'you-me-see'4, or by a compound of the type noun

1 Cf. Comment les cat6gories grammaticales changent de sens, to appear in theBulletin de la Soci6t6 de Linguistique de Paris.2 See Studien zu einer historischen Tempustheorie des Indogermanischen,forthcoming in Kuhns Zeitschrift fUrvergleichende Sprachforschung.3On the non-IE languages quoted, cf. F. Boas' Handbook; Kleinschmidt,Grammatik der gr6nlandischen Sprache, Berlin, 1851; Torrend, A ComparativeGrammarof the S. Afr. Bantu Languages, London, 1891;F. R. Blake, A Grammarof the Tagalog Language, New Haven, 1925; A. Dirr, Einfiihrung in das Stu-dium der kaukasischen Sprachen, Leipzig, 1928;S. H. Ray, Comparative Studyof the Melanesian Languages, Cambridge, 1926; v. d. Gabelentz, ChinesischeGrammatik 1881;Meillet-Cohen, Les langues du monde, 1924.4 This translation is of necessity incorrect since the personal prefixes do notexactly correspondto our nominative and accusative; cf. note 9.229


3/14

230 H. V. VELTEN+ verb as in Nahuatl ni-naka-kwa 'I-meat-eat'. If a verb-that isto say, a word or phrase expressing what would be regarded as a verbalidea in our IE languages-lacks an object or objective affix, it cannotbe distinguished from a noun or nominal phrase; cf. Greenlandic nan: oq'(it is) a bear', o:mavoq 'he lives', angivoq 'he is big'; Bantu mu-ntu,u-lede, omu-chle '(it is) the man, he sleeps, he (is) handsome', pluralba-ntu ba-lede aba-chle. The prefixes which determine the class of aBantu word join all words without exception and only transitive verbsare recognisable as a special grammatical category by a pronominalinfix as in mu-ntu u-ndi-bonide 'the man he-me-saw'. It must be em-phasized that we are concerned here solely with grammatical, not withlogical distinctions. A Chinook, e.g., may be quite well aware thatthere is a logical difference between the phrases bigness-of-him = 'he isbig' and house-of-him = '(it is) his house', but his language fails to dis-tinguish them. Furthermore, it should be kept in mind that many ofour most common grammatical terms are applicable only to IE lan-guages, and some of them only to the modern dialects. Thus the desig-nation transitive is used in this paper to characterize all verbs which re-quire a complement, i.e. the indication of the object-or the goal of theaction as in Latin eo domum, OE he eorfan gefeoll. For the term in-transitive we shall, therefore, substitute the more precise absolute.In the Malayo-Polynesian, the Sino-Tibetan, and a number of Cau-casian languages we find, instead of the distinction between absoluteand transitive phrases, a different type of division. Verbs which ex-press an action proper are distinguished by their grammatical form fromverbs designating a state or quality; compare, e.g., Tagalog ungmdralakd (nominative) 'having-taught-I' = 'I taught' with inibig ko (gen-etive) 'loving-of-me' = 'I love'; Georgic vaseneb'I build' with m-dzul-s'to-me-hate-is' = 'I hate'. In Chinese the living words,which expressan action caused by the will of the subject, are distinguishable from allother words by the tone with which they are pronounced. It is obviousthat in these languages the verbs designating a state, which would bestbe called affectiveverbs-as over against the actional verbs5-are, orformerly were, nominal forms. Consequently it may be conjectured

6 It is to be noted that our distinction of affective and actional verbs does notquite coincide with that of Tatverba and Empfindungsverba assumed by F. N.Finck (Kuhns Zeitschr. 41; Die Haupttypen des Sprachbaus, 1910), who usedthis division as the basis of his classification of languages as anreihende Sprachenand ankniipfende Sprachen-a principle of classification rendered obsolete bySapir's Language.


4/14

ORIGIN OF CATEGORIES OF VOICE AND ASPECT 231with some probability that the original nominal or absolute phrase,which ordinarily describes a state, gradually absorbed those transitiveverbs which do not express an action proper. In Greenlandic, whereno differentiation of affective and actional verbs is found, a form liketusarp-a-ra 'I hear him'6 represents a verbalphrase because it containsan object, while in Georgic the corresponding form mesmis 'to me-hear-ing-is', whether accompanied by a designation of the object or not,belongs to the affective-nominal class since it does not indicate an act ofvolition. Needless to say that the division of the nominal and verbalcategories may take, and in number of languages has taken, a quitedifferent course of development; witness the description of the verbsystem of Yana in E. Sapir's Language 111, 126.While originally a verbal idea could logically, in accordance with itsmeaning, appear only in one of the two classes, many actional verbslater developed also an affective form and vice versa, a phenomenon forwhich Georgic furnishes the most instructive examples. As J. Ven-dryes has pointed out,' the difference between Georgic msurs 'to-me-desire-is' and visurveb 'I desire' corresponds exactly fo that betweenthe impersonal8 and personal forms of the IE verb. The examples justquoted may be compared with the German forms mich verlangtand ichverlange. On the other hand, the distinction of the actional and affec-tive categories of the Georgic verb has come to express a difference ofaspect and tense, a process which is also parallelled in IE, as will beshown below.The above examples from Georgic and Tagalog indicate that thedifferentiation of the two verb classes is based on a distinction of case.The transitive type achievement-of-it-of-me, sight-of-it-of-me (double

6 It may of course be argued that tusarp-a-ra literally 'hearing-his-mine' isreally a nominal form. Again it must be rememberedthat the terms nominal andverbal are employed, not in their IE sense, but as terms of universal syntax.According to our above definition, any phrase that indicates both the startingpoint and the goal of a process or action represents a verbal expression. In otherwords, the distinction of the nominal and verbal categories is much less developedin many American and African languages, than it is in IE.7Cf. Le Langage, Paris, 1921,124.8 The term 'impersonal' is, of course, a misnomer since the verbs in questionappear in the form of the third person. But then the designation 'person' is alsoincorrect, and it is hardly possible to change a terminology which has been uni-versally employed for centuries; cf. O. Jespersen, The Philosophy of Grammar(London, 1924) 212. A complete bibliography concerning impersonal verbs willbe found in O. Behaghel's Deutsche Syntax 2. 120f.


5/14

232 H. V. VELTENpersonal affixes without case indication) was differentiated in languageslike Georgic by means of case forms as (a) actional achieving-it-I, (b)affective: (1) sight-of-it-to-me,or (2) seen-it-by-me. If some languageslike Basque employ the type b2 for all their transitive verbs, it is notpermissible to conclude, as Uhlenbeck and Schuchardt have done,' thatin all languages transitive verbs originally had the character of a 'pas-sive'. In IE just the opposite is the case. The IE active goes back tothe transitive-actional class, while the impersonal and the medio-passiveverbs, which cannot be separated as to their meaning, comprise onlysuch 'transitive' verbs-i.e. phrases expressing both the starting pointand the goal of a process-which can be classed as affective: e.g. Lat.me pudetalicuius rei, early NHG mir gelingt eines Weges.In IE three main classes of affective verbs are found:I. The absolute-affective group = verbs expressing necessity, obli-gation, propriety, etc.'-; cf. Skt. kdlpate 'it is in order, es gelingt', Lat.decet, oportet, Goth. daug 'es taugt', OHG zimit 'es ziemt', MHG esgebilrt,Pol. przystoi 'it is proper', OCSljestil + infinitive = Gr. laLn 'itis necessary', cf. Lat. est videre.Beside these impersonal verbs, medio-passive forms occur, especiallyin the modern dialects (see below); cf. Gr. 6iTo/a2 'ich bedarf' besidebZ 'es bedarf'. Such doublets are, of course, very rare since mostlanguages give preference to one form to the exclusion of the other.In Mod. Engl., e.g., there are hardly any impersonalia that could beplaced beside forms like I am obliged, I am bound, while impersonalverbs of the type of Pol. trzeba 'it is necessary', moina 'it is possible'abound in the Slavic languages, which have lost the IE medio-passive.That verbs of this class go back to nominal phrases is clearly indicatedby the usage of the Greek noun xp' as an impersonal verb in the mean-

9Cf. Uhlenbeck, Indog. Forsch. 12. 170,Kuhns Zeitschr. 39. 600;41. 400;Karak-ter k. bask. gramm. (Amsterdam, 1916)28; Schuchardt, Indog. Forsch. 18. 528;Berl. Akad. 1921. 651; compare the criticism offered by Finck, Kuhns Zeitschr.41. 209ff, and by Sapir, Internat. Journal of Am. Ling. 1. 85. Concerning adetailed treatment of the interrelationship of the categories of case, gender, andverbal affixesin this connection, cf. the study quoted in note 1.10We disregard verbs designating natural phenomena like Skt. varqati, Gr.,EL,

Lat. pluit, OCS1 dfiditfi, Goth. rigneib, Arm. telay, Tokh. B suwah, etc.'it rains', which are of necessity impersonal in all languages. For they indicateneither subject nor object; in other words, they are survivals of an original nom-inal category. Phrases expressing a subject like Jupiter tonat are doubtless of acomparatively recent origin and are to be attributed to the later preference forpersonal constructions. Cf. Siebs, Kuhns Zeitschr. 43. 253ff.


6/14

ORIGIN OF CATEGORIES OF VOICE AND ASPECT 233ing 'it is necessary' and by an expression like apKET6rby which cannotbe distinguished as to its form from a nominal phrase like

rTKvoVUoL;

in the Slavic languages nouns are quite commonly employed in thismanner: cf. Pol. czego '(it is) a pity', etc.II. Absolute forms of actional verbs, i.e. the impersonal, or rather theindefinite, use of the third person of ordinary personal verbs; cf. Lat.convenit, Goth. gaqimib 'it is proper', and the Old Lat. legal phrase,found in the Law of the twelve Tables, si in ius vocat,ni it, antestamino'if (one) calls (somebody) to court and he does not come, a calling ofwitnesses is to take place', which according to Wackernagel" representsan instructive example of Latin syntax of the fifth century B. C.The same usage occurs in Greek and is very common in Keltic; cf.OvyaTpl 'v L6


7/14

234 H. V. VELTEN]aurseib mik 'mich dtirstet', Lat. me pudet, me paenitet, me taedet. InGreek and Sanskrit, i.e. in the only languages which have kept the IEmedio-passive intact and have developed it into an elaborate system,impersonalia of this group are almost completely lacking and medio-passive forms are used instead;'3 cf. Gr. 'uat, Skt. d'ste 'sit', Gr. KdEaL,Skt. fite 'lie', Skt. sevate 'remain', 6dhate'thrive, rdmbate, Gr. Kp pajlat'hang', Skt. jdrate 'rustle', vepate 'tremble, oscillate', vrddate 'becomesoft', Gr. KLvvUaLbe in motion', airoyat 'rot', rcpaoEoaLbecome dry',aAbXoyat 'burn' (intr.), O8poiat 'become warm', KLalVaaL, Skt. prd-thate 'spread' (intr.), EpvOaLvoliatbecome red'; and the verbs pertain-ing to a mental state: Skt. mrdate

'be glad', bla'ghate'be confident', Gr.XIo L, aKVotroLaL'be angry', xvvtiat 'be sorry', &'oiat, -c'oiat 're-vere', 6aaoj.at 'ahnen', fobiAoia, e~Xoi0aL 'wish', /IatooiaL 'long for',epaIaLt'love', Opda'o/iat 'imagine, think over', OKcarro/atbe looking at,contemplate'. Most of these verbs are media tantum, while a few ofthem developed active forms-usually of a causative-transitive meaning-in the later period of the language.14The last nine examples show clearly that the verbs of this group can-not be simply classed as intransitive, but that they form an affectivecategory as it is found in Georgic and other non-IE languages. Thatthe medio-passive forms have replaced older impersonalia in Greek andSanskrit is made probable by the occurrence of the isolated HomericAoi blEraT (r 312) 'mir ahnt' beside the personal bLoiat 'ich ahne',and such impers. passive forms in Skt. as supyate tvaya 'es wird von dirgeschlafen', which may be derived from older affective-nominal forms:'premonition (sleep) is to me'. Furthermore, the Vedic examples, ki-tavdrm atapa (RV 10. 34. 11) 'es schmerzte den Spieler', nd md taman ndsraman ndta tandrat (RV 2. 30. 7) 'may not inertia, fatigue, or lassitudeseize me', surely represent quite archaic constructions."The medio-passive is evidently of more recent origin than the affect-ive type grief-of-me, atigue-in-me, thought-mine,which is found all overthe earth and occurs in the most heterogeneous languages. In IE therearose then a tendency, first to generalize the verbal construction and tochange such phrases to it grieves (within) me, there s thoughtin me, me

13 Armenian also employes medio-passive verbs to express affective meanings;cf. gtham'I have pity', yowsam I hope'.14Cf. Delbriick, Grundriss, 4. 417.15In place of the impersonal tandrat, the personal mediopassive tandrayate sfound in AB; there is little plausibility in Delbriick's assumption (Grundriss5.30) that the personal form is older.


8/14

ORIGIN OF CATEGORIES OF VOICE AND ASPECT 235thinks, and then to extend the personal form of the actional verb byturning the affective class into something like I grievewithin myself, Ihave thought (fatigue, etc.) in myself. This does not mean that imper-sonal constructions were transformed abruptly into medio-passiveforms. Most probably impersonalia first changed to personal verbsmodeled on the actional category; but since such affective verbs werestill felt to differ in meaning, they came to be distinguished by a fullerpersonal ending. As Brugmann has pointed out,16 the medio-passivesuffixes *-tai, *-toetc. are in ablaut relation to the active endings. Thisinterchange may have taken its starting point in the affective forms oforiginally actional verbs. For it is well known that the medio-passiveregularly served to turn active-transitive verbs into intransitive ones,17cf. 0alvw 'I show', calvouaL 'I appear', veho'I drive', vehor ich fahre',rumpere 'break' (trans.), intr. rumpi.Except for Slavonic, which has not only kept intact the impersonalclass but also preserved the older category of verb aspects in preferenceto the more recently developed principle of tense distinction, almost allIE dialects tend to abandon impersonal constructions.As regards Germanic, there is no proof for Delbriick's contention"sthat impersonalia like Goth. huggreibmik, baurseib mik, which are mostprobably denominal verbs, go back to personal forms meaning originally'ich habe Hunger (Durst)'. It must be objected first of all that theidea 'to have', and the corresponding verb type habeo, exw,etc., is quiteyoung; in most non-IE languages as well as in the older IE dialects, andto this day in Slavic, the older type mihi est, or- /ot~ prevails: cf.Skt. ddha sma te vrdjanam?r~~ymasti 'dann ist dir die Bahn schwarz',Goth. ni was im barne (Luke 1.7) 'obK Vpabro^s rKVOV'.In the modernGermanic languages there is a marked increase of personal forms.Thus Mod. Engl. discarded me thinks, me likes, me pleases, etc. in favorof personal constructions; OHG mir anget, mir unmahtithave given wayto NHG ich dngstige mich, ich werdeohnmdchtig,and the NHG medio-reflexive forms (see below) ich wunderemich, ich ergrimmemichdariiber,ich bange mich, ich graue mich, ich ekelemich are much more commonthan the corresponding impersonalia es wundert, ergrimmt,mich, mirbangt,mir graut, es ekeltmir.'9 It is to be noted that this development

16Griechische Grammatik3 10f.17Cf. F. Sommer, Vergleichende Syntax der Schulsprachen47.18Grundriss 5. 33; followed by Brugmann,Kurze vergl. Gramm. 629.19 Needless to say that isolated examples of a change in the opposite directioncan be found. Active (non-reflexive) forms are disregarded for the present; videinfra.


9/14

236 H. V. VELTENtook place after a new analytical passive had been created, and thatthe tendency to get rid of impersonalia is most pronounced in thelanguage which uses the passive most frequently, viz. in Mod. Engl.;compare OE me sceamabwith I am ashamed,NHG mir ist schlecht,es tutmir leid with I am ill, I am sorry.20In Latin affective verbs of the impersonal type like me piget, me taedet,me miseret, mihi lubet (libet) are found side by side with the medio-passive type misereor, laetor, meditor, reor, the personal constructiontending to be preferred;cf. note 19. In the latest period of the language,however, when the r-passive was on the point of disappearing, imper-sonalia seem to increase again; cf. the impersonal horret 'mir graut',debet n the sense of 'necesse est', and habet, used like Fr. il y a.21 TheRomance languages, finally, manifest the same tendency as the modernGermanic dialects: after a new method of expressing the medio-passivehas arisen, the impersonal forms are gradually abandoned; cf. je mesouviens,je regrette or older il me souvient,il me regrette.As nearly all actional verbs developed medio-passive forms, the medio-passive came to indicate quite generally any action which takes placewithin the sphere of, or in the interest of, the subject. This was sensedvery accurately by the Hindu grammarians who called this diathesisdtmane padam 'word for oneself'. The usage is illustrated by suchexpressions as Skt. ydjate, Gr. OberaLhe sacrifices for himself', as overagainst ydjati, ObELhe sacrifices for another' (said of the priest). Thusthe medio-passive includes not only the so-called 'dynamic' middlevoice22comprising such phrases as -roXiov apxe&Oat 'to undertake awar in one's own interest or under one's own leadership', in contrast to

20 According to the generally accepted explanation, forms like him likes werereplaced by he likes because of the loss of the nominal inflection, since the noun intheking likes 'dem Konig geffillt es' could no longer be recognized as a dative andwas mistaken for the subject case. The passive extended its sphere because, ow-ing to the lack of accusative forms, a sentence of the type of NHG meinen Sohnhat er getatetis usually expressed in Engl. in the form my son was killed by him.It would, however, be wrong to state that the loss of the noun inflection broughtabout the disappearance of impersonal verbs and the increase of the passive con-struction. For as Htibener (Beitr. z. Gesch. d. deutsch. Spr. u. Lit. 45. 85ff.)has shown conclusively, the reverse process took place: the preference for thefixed word order subject-verb-object (which implies a preference for personalconstructions beginning with the subject) caused the gradual abandoning of theOE case forms.21Cf. Wackernagel, Vorlesungen 1. 117-9.22 A somewhat vague term, which Wackernagel (Vorlesungen 1. 127) is fullyjustified in rejecting. However, he does not suggest a substitute.


10/14

ORIGIN OF CATEGORIES OF VOICE AND ASPECT 237roM~ioviapXtv 'to cause a war', but also reflexive verbs like Xov'ojaL,lavor 'I wash myself', ponor 'setze mich' (Cicero, Epist. 9. 15. 4). Theresult was a more or less complete superimposition of the reflexive andmedio-passive categories: one has the choice of saying rIapaoWKEVELViav-ry, 7yvTv&6rv avrYv, se exercere, se lavare or rapaaKEv&UEo6OaL,yviYvEeaGat,exerceri, lavari.The close connection of the medio-passive with the reflexive verb hasbeen noted in all grammars since the times of Dionysios Thrax. Inmany languages the reflexive gradually took over some or all of thefunctions of the medio-passive. Among the IE dialects this processwas carried farthest in Slavic, Lettic, and Scandinavian;23 it is foundalso in German-cf. Otfr. 4. 32. 2, s6 spialtun sih thie steina 'so spaltetensich die Steine', NHG das versteht ich vonselbst,das lernt sich leicht, dasvergisstsich bald-as well as in Late Latin: e.g. Myrinam quae Sebasto-polim se vocat (Plinius, Nat. Hist., 121), facit se hora quinta (peregr.Aether. 27. 3), and in Romance: Fr. cela se voit, se comprend,etc. Asto non-IE languages, the same phenomenon occurs in Semitic, wherethe old passive of the vowel sequence a-a-a (Arabic u-i-a) has beencrowded out partly by reflexive forms, notably in Arabic and Ethiopian;cf. Eth.

tant.le'a 'he has been covered'. In Finno-Ugrian and in Turk-ish the passive is identical with the reflexive; cf. Finn. murtu- 'se briser,be broken' (murta- 'break' trans.), Turk. a3-yl-mak 'sich 6ffnen, beopened' (ai-mak 'open' trans.).To a certain extent the reflexive serves in the modern Germanic andRomance languages to express the middle voice as distinguished fromthe passive proper.24 However, the coexistence of a pure passive of thetype filius a patre laudatur, Peter is beatenby Paul and an active formpaterfilium laudat etc. is comparatively rare and represents, whereverit is found, a quite late formation, as Wundt has pointed out.25 Orig-inally every passive combines the functions of what the Greek grammarscall the middle voice and the passive. Even in Greek and Sanskrit-the only languages to distinguish three diatheses-the formal differ-entiation was never carried out consistently.26Additional proof for the impersonal and nominal character of the

23 Cf. Pedersen, Kuhns Zeitschr. 40. 160ff.24 Cf. Sommer 51, Streitberg, Gotisches Elementarbuch3-4 189, Behaghel,Deutsche Syntax 2. 199.25 On this question cf. Stolz-Schmalz, Lateinische Grammatik4(Munich, 1910)490f. and the literature quoted there.26 Cf. Delbrtick, Grundriss4. 432ff; Wackernagel 135ff.


11/14

238 H. V. VELTENmedio-passive is furnished by the universal observation, holding goodfor the most divergent languages, that passive sentences are used morefrequently without the agent,27and that more often than in all othertenses the passive appears in the perfect, especially in the perfect parti-ciple-that is to say, in a non-actional form.28 Thus the type filius apatre laudatur is rarer than the simple filius laudatur, while laudatur isless frequent than laudatus est.29 Passive participial constructions arefound also in Sanskrit and Attic; cf. igtd'd8vdtah'revered (are) the gods'(occurs also with the copula), -yeypauIuEvo0LEl. In all IE dialectsnominal forms of the passive survive longest; witness the IE verbaladjective in *-to-,which is preserved even in Slavic (OCSIz tu 'hewn' =Skt. hatds 'beaten'), in Lithuanian (de'tas 'placed' = Skt. dhitds, Gr.oErbs),and in Italo-Keltic (M. Ir. dobreth datum est', Skt. bhrtds 'car-ried').29aHowever, the identity of the medio-passive diathesis and the affectivecategory has become effaced to a certain degree even in the oldest IEdialects since, as it has been shown above, the medio-passive came todesignate every action carried out in the interest of the subject; thus thetype 7rotoiad 'I do for myself' could no longer be classed as affective.On the other hand, beside many of the affective verbs quoted above ingroup III, active forms of the same meaning could be adduced, e.g.AX&E'eside gpauat, &KoVW,&yardw, 7rCu4weside Ko0bouaL, AfarbaolaL,r~uEojaL;pudeo, paeniteo beside me pudet, me paenitet, gaudeo besidelaetor; and many active Skt. verbs in -ya- like va'yati 'become tired',pftyati 'thrive', md'dyati 'be glad' beside the medio-passive and imper-sonal forms given above. This gradual loss of the original significanceof the active, which could at first express only actions proper, is to beattributed to the fact that the grammatical distinction of the actionaland affective classes engendered, not only the category of voice, butalso the differentiation of the verb aspects.The original type of the affective category sight-to-me(of-me) can beinterpreted in two ways, namely (1) sight (an image) is in me or (2)an image strikes me. It is obvious that the former interpretation leadsto the origin of the durative aspect (1), while the latter represents the

27See Stolz-Schmalz, loc. cit., the literature quoted by Wackernagel (143f.),and L. G. Frary, Studies in the Syntax of the Old English Passive (LanguageDissertation No. V, 1929)7.28Cf. Sommer 49.29 Cf. Stolz-Schmalz. 2546.29aCf. the Germanicand the Romance passives formedfromperf. participles.


12/14

ORIGIN OF CATEGORIES OF VOICE AND ASPECT 239instantaneous (perfective) action (2) as soon as the idea of duration isassociated with the verb system. In Georgic the affective class seemsto have been identified with the aspect (2)30-compare, for example, theactive-durative present vxedav 'I see' with the affective-instantaneousaorist dedamdainaxa mt'xowara'to the mother (casus activus) appeareda beggar'-whereas the opposite happened in IE: cf. bipKogtat,CspatKOv,revertor,reverti, assentior, assensi. The alternation between the medio-passive present UpKoLaL 'blicken, be looking' (e.g. 6Ebv ~ipKEcTOaL'furchtbar dreinblicken') and the active aorist 'paiKov 'ich erblickte, Inoticed', which represents a survival of a quite archaic type of con-jugation never satisfactorily explained hitherto,3' obviously indicates adifference of aspect, not of diathesis.32The actional verb class, classified as to duration, also becomes dividedinto two groups, namely: the verbs of the terminative aspect (3), ex-pressing an action directed toward an aim, build, fight, strike, endeavor,and those verbs which imply the attainment of a goal like reach, achieve,defeat,kill. It goes without saying that the latter group is identical, asregards its aspect, with the perfective verbs derived from the affectiveclass. Thus we obtain the following three categories.

(1) durative (2) instantaneous (3) terminativeaKi~rro7a&be looking' l&elvnotice' 6pcic'observe' ('aim tosee')9PXOlsat 'be marching' 13ivac, XOdivarrive' T7p~xo run' (toward agoal)All three aspects are characterized by the personal suffixes, and itappears that the so-called secondary endings constitute the earliest man-ner of designating the aspect (2). The original form of this aspect mustthen have been closely akin to, if not identical with, the injunctive. Itwas not until a later period, after the aspect (2) had come to represent a30 It is interesting to note that a crossing of the categories of voice and aspectin the same direction has taken place in French, where the medio-reflexive verboften serves to express the perfective aspect, e.g. in se ddfiler,se trotter;cf. Ven-dryes, Le Langage 131.31Cf. Wackernagel 134.32Active perfect forms with an intransitive meaning-often used as a presentlike b~EopxaI see'-occur commonly beside medio-passive presents; cf. ykyova-"yly'vota,a ovXa-ftobojaa; the same interchange is found in Keltic; concerningSkt. examples see Bloomfield-Edgerton, Vedic Variants 1. ??45, 69. This is notsurprising since the perfect itself is of durative character and serves as a deviceof durativation; cf. Sommer 66. The perfect with medio-passive endings is ofyounger origin; see Delbriick, Grundr.4. 415. Concerning an explanation of themedio-passive future of active verbs cf. Delbrtick, Synt. Forsch. 4. 74.


13/14

240 H. V. VELTENpast tense, i.e. the aorist, and after a durative past, the imperfect,hadbeen created by means of the augment, that the secondary endings losttheir perfective meaning and took on a temporal, and later a modalsense.33In the earliest period of the category of aspect the personal endingsserved as a method of durativation, perfectivation, etc. Of this statethere remain, however, only a few traces in historical times; cf., besidethe above examples, dy,,XXopa ,qyXat, yvKaoyatLevKo, aiLVo/LaLMuany. s soon as the tenses become established, medio-passive aoristsappear beside the older active forms, an active present is formed ofmany durative-affective verbsm, and vice-versa. In other words, theubiquitous tendency to level out forms makes itself felt, and by a cross-ing of the categories of voice and aspect we obtain four forms in placeof the former two or three: roLz-o,otLoiIaL, roloca,tror)soyrv. Mean-while new grammatical devices for designating the aspect had beencreated-notably the *io augmentation of the durative-present stem,the aorist *s, the perfective prefix, and the perfect already referred to-while the aspect (3) was gradually absorbed by the other two. Thisdevelopment and the further evolution of the categories of aspect andtense have been traced in detail in the paper quoted in note 2.Thus the medio-passive loses to a large degree its durative as well asits original affective character. In the oldest Vedic literature we canclearly perceive a period of transition: the distinction of the active andthe middle voice has become effaced to such an extent that the forms areoften interchanged at random.35 In the later language, therefore, theyounger (present) passive in -yd- gains over the older medio-passiveforms, which are characterized by the personal endings alone. Tracesof the durative nature of the passive, however, have survived in all IEdialects; above we referred to the relative frequency and persistency ofpassive forms in the perfect-i.e., in the tense which developed fromdurativated verb forms.The syntactical evolution of the IE voices and aspects traced in thisstudy may then be summed up as follows.A. Division according to complements:(1) absolute-nominal phrases (2) transitive or objective phrasesType: house-mine, bigness-his, Type: doing-of-it-of-him,thought-yours. perception-its-mine.Gr &pKE6V/IOL,7rKVV 1OL. Lat. cura mihi alicuius rei (est)13The aorist augment is derived from the imperfect; cf. note 2.

34Often with a causative meaning; cf. A~dAXXcglorify', which appears in thelater language beside the Homeric A-ydXXogauboast'.3 See Bloomfield-Edgerton, Vedic Variants 1. 30, 33ff.


14/14

ORIGIN OF CATEGORIES OF VOICE AND ASPECT 241B. A rigorous distinction of nouns and verbs develops, verbal suffixesappearing even in phrases that indicate neither subject nor object:

Skt. tdpati 'it is warm', Gr. v?ite, 36e, etc. At the same time a moreintricate case system is formed, permitting of a differentiation of thefollowing classes.(1) impersonal-affective (2) personal-actional verbsmich hungert, me thinks, domum eo,il m'apparaUt,me misereteius. domum construo.C. The personal construction encroaches upon the impersonal, thedistinction of the affective and actional classes being preserved, however,by means of a new grammatical device. Thus we obtain:(1) affective-medio-passive (2) actional = active verbs

idhate, epatat, meditor. EKELKYUt,Ot0,rp70.Newly developed medio-passive forms of active verbs like Xobosat, avorbring about the partial identification of the medio-passive with the re-flexive verb.D. The idea of duration is associated with the verb system. Ac-cordingly, the two classes are subdivided and partially superimposed,a process from which three new categories result:(1) durative (2) perfective (3) terminative verbs(medio-passive endings) (secondary endings) (primary endings)For a continuation of this synopsis see the final table in my study citedabove.

on the origins of categories of voice and aspect

Documents