71
--------------------------------------, JAY H. JASANOFF STATIVE AND MIDDLE IN INDO-EUROPEAN INNSBRUCK 1978

Jasanoff.1978.Stative and Middle in Indo-European

Embed Size (px)

Citation preview

Page 1: Jasanoff.1978.Stative and Middle in Indo-European

--------------------------------------,

JAY H. JASANOFF

STATIVE AND MIDDLE IN INDO-EUROPEAN

INNSBRUCK 1978

Page 2: Jasanoff.1978.Stative and Middle in Indo-European

Die INNSBRUCKER BEITRAGE ZUR SPRACHWISSENSCHAFT werden gefiirdert durch das Bundesministerium für Wissenschaft und Forschung in Wien, die Kulturabteilung

des Amtes der Tirolel' Landesregierung und durch den Universitatsbund Innsbruck

Die Publikation der vorliegenden Studie erfolgte mit zusatzlicher Unterstützung des Fonds zur Fiil'derung del' wissenschaftlichen Forschung, des Verbandes del' wissenschaftlichen Gesellschaften bsterreichs (auf Antrag der fnnsbrucker Sprachwissenschaftlichen Gesellschaft)

und des Department of Linguistica der Harvard University

Adresse des A1ttorS: Professor Jay H. Jasanoff

Depal'tment of Modern Languages and Linguistics Coruell University

Ithaca, New York 14853, USA

ISBN 3-85 124-540-7

1978

INNSBRUCKER BEITRAGE ZUR SPRACHWISSENSCHAFT Herausgeber: Prof. Dr. Wolfgang Meid

Institut für Sprachwissenschaft der Universitat Innsbruck A-6020 Innabruck, Innrain 30

Druck: Ernst Becvar, A-1150 Wien, Lichtgasse 10

ACKNOWLEDGEMENT

This book owes much to the advice and criticism ofmy friends, colleagues and students at Harvard University, to whom 1 am happy to express my gratitude. Thanks are due also to the Editor, Prof. Dr. Meid, for his assistance in pteparing the manuscript for the press, and to the Depart­ment ofLinguistics, Harvard University, for helping to defraypublication costs.

Cambridge, Mass. June, 1978

5

Page 3: Jasanoff.1978.Stative and Middle in Indo-European

CONTENTS

Bibliographical abbreviations

1. Introduction: The Stative Formations oí Indo-European .

§ 1. Deverba tive and denomina tive sta tives. p. 13. § 2. The perfect: form and function. p. 14. § 3. The perfect and middle endings. p. 15. § 4. Expansion ofthe perfect middle. p.15. § 5. The suffix *-e-. p. 16. § 6. Denominative e:. statives. p. 17. § 7. "Primary" e-statives. p.18. § 8. The problem of the distribution of *-e-. p.19. §9. The type *mineti, *minT-: the semithematic theory. p.19. §10. *mineti: other views. p.21. §11. The Germanic type *habaiP: competing theories. p.21. § 12. The putative alternation *-e(i)-: *-~(i)-. p.22. § 13. Conclusions. p.23.

9

13

II. Class III Presents in Tocharian . . . . . . . . . . . . . . . . .. 24

§ 14. Present of AB mask-. p. 24. § 15. List ofverbs with class nI presents. p.25. § 16. Class nI subjunctives. p.25. § 17. Present of AB plant-. p.26. § 18. List of class IV presents. p. 27. § 19. Class nI presents and e­statives. p. 27. § 20. Inadequacy 'of the proposed derivation *-a- < *-e-. p.28. §21. Phonological source ofthe class nI theme vowel. p.29. §22. Absence of distinctive quantity in Tocharian. p. 29. § 23. Phonetic interpre­tation of the Tocharian vowels. p.30. §24. The Pre-Tocharian vowel system. p. 31. § 25. Primary rounding and uurounding. p. 32. § 26. Later developmentsin Toch. AandB. p. 33. §27. Phonologicalhistory ofAasatar, B osotar. p. 34. § 28. Syncope of -a- in Toch. A. p. 35. § 29. "Persistent" *-0-. p.35. § 30. Class nI subjunctives and thematic presents. p. 36. § 31. Class nI (IV) roots in -sk-. p. 37. § 32. Root-final -tk-. p. 38. § 33. Class nI (IV) presents and lE aorists. p. 39. § 34. B lip:'; wak-. p. 41. § 35. Morphological idiosyncrasies of the thematic aoristo p. 41. § 36. Development of class nI presents frommiddle root aorists. p. 42. §37. Other roots: AB as-, B palk-, etc. p.43. § 38. Apophonically deviant forms: the types A tsarwatar, B ñewetar. p.44. §39. Class nI forms with active endings. p.45. §40. Smnmary. p.46.

Appendix: The Thematic Middle in Tocharian and Indo-European 47

§4L Persistent *-0- elsewhere in lE. p.47. §42. The thematic vowel in Balto-Slavic. p.48. § 43. The thematic vowel in Hittite. p.49. § 44. The thematic conjugation in Indo-European. p.50. §45. Introduction of the thematic middle. p.51. §46. Later remodelings of the thematic middle. p.52. § 47. Original distribution of the endings *-e and *-0. p. 53.

7

Page 4: Jasanoff.1978.Stative and Middle in Indo-European

III. The Third Weak Class in Germanio. . . . . . . . . . . . . . .. 56 §48. Inflection of 'have' in Germanic: paradigms. p. 56. §49. Irregularities and special features. p. 56. § 50. Synchronio position of the third weak class. p.58. § 51. The stems *halJai-, *haba-, *habja-. p. 59. § 52. Earlier theories of the class In suffix: *-e(-iefo)-. p.60. §53. Earlier theories (cont.): *-ei-: *-t- (*-¡¡i-). p.62. §54. The Hittite type déii: tiyanzi. p.63. §55. Earlier theories (con t.): *-¡¡-iefo-. p.64. § 56. Apophonic behavior of lE *-e-(*-ehr ). p. 66. § 57. The problem of *-a- vs. *-ja-. p. 67. § 58. Distribu-tion of *sayja-. p. 68. § 59. Distribution of *habja-. p. 69. § 60. Distribution of *libja-. p. 70. § 61. The CGmc. paradigm. p. 71. § 62. Formal renewal of the 3 sg.: *rfmip, *bibaip. p. 71. § 63. Replacement of *-ai by *-aip in the 3 sg. middle. p.73. §64. Morphological behavior of *dheuyh- in Indo-Euro-pean. p. 73. § 65. Gmc. *duyaip and Ved. duhé: development of the class nI paradigm. p. 74. § 66. Gmc. *fulyaip and Ved. sp'(sáti. p. 76. § 67. Class nI verbs beside preterito-presents. p. 77. §68. Gmc. *m~maip and Av. mamne. p. 77. § 69. Gmc. *witaip and Ved. vidé. p.79. § 70. Gmc. *Parbaip, *mayaip. p.80. § 71. Perfect middles and root aorists. p.80. § 72. Gmc. *libaiP,' *wunaip, *pulaip; status of *hanyaip. p. 82. §73. The type *kun-naiP. p. 84. § 74. Class nI denominatives. p. 86. § 75. Modaland particip-ial forms. p.88. § 76. Runic witadahalaiban and Ved. vidádvasu-. p.89. § 77. The preterite ofthe third weak class. p. 90. § 78. Formal history of the Gmc. dental preterite. p. 91. §79. Summary. p. 93.

IV. Stative Presents in *-t- in Balto-Slavic ..... . § 80. Inflection of *mintti in Balto-Slavic: paradigms. p.94. § 81. Earlier analyses: introduction. p.95. § 82. The semithematic theory. p. 96. § 83. The contraction theory. p.97. §84. The ablaut theory. p.98. §85. The theory of Kurylowicz. p. 100. § 86. Other i-presents in Slavic. p. 101. § 87. The Baltic s-future: form of the tense signo p. 103. § 88. Lith. dúosiant- and Ved. déisyánt-. p.l04. § 89. Extension of -i- from the 3 pI.: OPr. waidimai, -ti. p.l05. §90. Spread of -i- in other categories: Lith. devi-, yari-, veizdi-. p. 107. § 91. The 3 pI. as starting point for the stative paradigm. p.l09. § 92. The perfect middle of *men- in pre-Balto-Slavic. p.l09. § 93. Establishment of the stem *mini-. p.110. § 94. Replacement of *mini- by *min;¿:" in Slavic. p.112. §95. Comparison oH-presents with middle forms elsewhere. p.113. § 96. OCS shcati, s'hpati. p.114. § 97. Miscellaneous other verbs. p.114. §98. Participial andmodal forms. p.115. §99. Summary. p.117.

V. Conclusion: The Role oí *-e- in Indo-European . § 100. Middle aorist presents: reappraisal of Ved. duhé. p.118. § 101. Relationship of the type *(me)mnór to *sikór and *dhuyhór. p.119. § 102. Uses of *-e- in the attested languages. p.120. § 103. Uses of Ved. yúhéi. p.122. § 104. Development of a finite conjugation from the predicative use of the instrumental. p. 123. § 105. Extension of*-e- (*ehr) as a verbal suffix. p.125.

94

118

Index oí Forms Cited . . . . . . . . . 127

8

BIBLIOGRAPHICAL ABBREVIATIONS

AfdA: Archiv für deutsches Altel'tum. Bartholomae: C. B., Altiranisches Worterbuch (1904). Bartholomae, Stud. Z. idy. Sprachyesch.: C. B., Studien zul' indoyermanischen

Sprachyeschich(e n (1891). BB: Bezzenbéryel's Beitl'ii{¡e. Benveniste Probo de liny. yén.: E. B., Pl'oblemes de linyuistique yénérale (1966). Brinkmann', Spl'achwandel: H. B., Sprachwandel und Sprachbeweyunyen in althoch-

deutscher Zeit (1931). Buck, Gram. ofOsc. and Umbr.: C. D. B., A GrammarofOscanar:d Um?1'ian (1904~. Brugmann, G1'. 2 : K. B., G1'und1'ij3 der veryleichenden Grammattk del' tndoyermant-

schen Sprachen 2 (1897-1916). .. . Brugmann, Gr(undrij3) 1 : K. B. and B. Delbruck, Grundrij3 del' veryleichenden

Grammatik der indoyel'manischen Sprachen 1 (1886-1900). BSL: Bulletin de la Société de Linyuistique de Paris. Chantraine, Gram. ham.: P. C., Grammaire homéTiq'ue I (1948). Ohantraine, Hist. du parf. yr.: P. O" Histoire du parfait yl'ec (1927). . Oollitz, Das schwache Praeteritum: H. O., Das schwache Pl'aeterttum und setne

Vo1'yeschichte (1912). Delbrück, Veryl. Synt.: see Brugmann, Gl'undrij31. Endzelin, Lett. Gr.: J, E., Lettische Grammatik (1923). Festyruj3 Roth: Festyr~lj3 an Rudolf von Roth ~um Doktol'-Ju?i1aum: (1893). FestscM. Pokorny: Beitrii{¡e zul' lndoyel'manisttk und Keltoloyte, J ultus Pokorny zum

80. Geburtstay yewidmet (1967). Flasdieck: H. F., Untel'suchunyen über die yel'manischen schwachen Vel'ben 111.

Klasse (= Anylia 59) (1935). Flexion und Wol'tbilduny: H. Rix, ed., Flexion und Wortbilduny, Akten del' V.

F achtayuny del' 1 ndoyermanischen Gesellscl~aft (1975). . Fraenkel Malchel' Pietkiewicz: E. F" Sprachltche, besonders syntakttsche Untersu­

chung des kalvinistischen litauischen Katechi81nus des Malcher Pietkiewicz von 1598 (1974).

Gedenkschr. Brandenstein: Studien zur Sprachwissenschaft und K ~lltul'k1tnde; Ge­denkschl'ift jül' W ilhelm Brandenstein (1968),

Geldner: K. F. G., Del' Riyveda (Harvard Oriental Series 33-35), Godel, lntrod. to the Study ofOlass. Arm.: R. G., lntl'od~lCtion to the St~ldy ofOlassical

Armenian (1973). ' Grassmann, Wb.: H. G., Wol'tel'buch Z1lm Riyveda. , Guxman, Srav, yram. yel'm.jaz.: M. G. et al" Sravnitelnaja yrammatika yel'mansktx

jazykov (1966). Holthausen, Alts. EI.2: F. H" Altsachsisches Elementarbuch 2 (1921). lnt. Jo~wn. of Slav. Liny. and Poet.: lnternational Journal of Slavic Linyuistics and Poetics .

9

Page 5: Jasanoff.1978.Stative and Middle in Indo-European

IF: Indogermanische Forsch-ungen. J AOS: J oumal of the American Oriental Society. Johannson, De derivo verb.: K. F. J., De derivatis vel'bis contractis linguae graecae

quaestiones (1886). Journ. of lE Stud.: Joumal of Indo-EuTOpean Studies.

Krahe, Germ. Sprachw.: H. K., Germanische Sprachwissenschaft lI6 (1967). Krause, WToch. Gram.: W. K., Westtocharische Grammatik 1 (1952). Krause-Thomas, Toch. Elem.: W. K. and W. T., Tocharisches Elementarbuch I

(1960), II (1964). Kurylowicz, Infl. Cat.: J. K., The Inflectional Categories of Indo-E7tTOpean (1964). Kurylowicz, (L')apoph.: J. K., L'ClIpophonie en indo-ew'opéen (1956). KZ: (Kuhns) Zeitschrift für vergleichende Sprachforsch7mg. Lg.: Language. Ling.: Slavistiéna Revija-Linguistiw. Mahlow, Die langen V okale: G. H. M., Die langen Vokale A E O in den eUl'opaischen

Sprachen (1879). Meid, Das germ. Praet.: W. M., Das germanische Praeterit7tm (1971). Meillet, Slav. com.: A. M. and A. Vaillant, Le slave comm7m2 (1934). JYI SS: JYI ünchener Studien Z7tr Sprachwissenschaft. N arten, sigm. Aor. : J. N., Die sigmatischen Aoriste im Vedc¿ (1964). Noreen, Altn01·d. Gl'. 1: A. N., Altnordische Grammatik 14 (1923). NTS: Norsk Tidsskrift for Sprogvidenskap. P BB : Beitrage Z7¿r Geschichte del' deutschen Sprache und Literat7M'. Pedersen, Toch. 2 : H. P., Tocharisch vom Gesichtspunkt del' indoe'uropaischen

Sprachvergleichung 2 (1941). Pokorny, IEW: J. P., Indogermanisches etymologisches Worterbuch 1 (1959). Pu~vel, Lar. and the lE Vb.: J. P., !-,aryngeals anel the Indo-E7tropean Verb (1960). RelChelt, Awest. El.: H. R, Awest'!sches Elementarbuch (1909). Reno~, Val. el7¿ palj. : L. R, L,a valeur d7¿ paljait dans les hymnes védiques (1925). Schmld, (Stud. z.) balto (7¿. ~dg.) Vb.: W. P. S., Studien Z7¿m baltischen 7tnd

inelogermanischen Verb7tm (1963). Schwyzer, Gr. Gram.: E. S., Griechische Grammatik 1 (1938). Skutsch, Kl. Schr.: F. S., Kleine Schriften (1914). Slav. and E. Et¿r. J01tl·.: Slcwic and East E7t1'Opean Journal. Sommel', Krit. E1'l.: F. S., Iúitische Erlautentngen Z7tr lateinischen Lmd- ttnel

F01'1nenlehre (1914). Spr. : Die Sprache. Stang, Sl. U. balto Vb.: C. S. S., Das slavische unel baltische Vel'bum (1942). Stang, Vergl. Gr. d. balto Spr.: C. S. S., Vergleichende Grammatik elel' baltischen

Sprachen (1966). Stl'eitberg, Z1tr germ. Sprachgesch.: W. S., Z1t1· germanischen Sprachgeschichte

(1892). Stl'ei~berg, Urgermanische Grammatik: W. S., Urgermanische Grammatik (1896). Stud'¿es Lane: St1¿dies in Historiwl Ling7tÍl3tics in Honor of George Sherman Lane

(1967). Thurneysen, OIGr.: R T., A Grammar of OlelIrish (1946). Toch. Spmchr. 1 : E. Sieg and W. Siegling, TocharischeSprachreste, SpracheB, Heft 1

(1949). TPS: l'ransactions of the Philological Society.

10

Uljanov, Osrwvy: G. K. U., Znaéenija glagol'nyx osnov v litovskoslovjanskom jazyke (1891 ).

Vaillant, Gram. comp. lII: A. V., Grammaire comparée deslang7tes slaves III (1966). Wackernagel, Kl. Schr.: J. W., Kleine Schriften (1955). W ackernagel-DebrUllller, Altinel. G1'. : J. W. and A. D., Altinclische Grammatik I­

III (1896-1957). Wagner, e-Yerba: H. Wagner, Zur Herkunft der e-Yerba in den indogermanischen

Sprachen (1950). Watkins, Celt. Vb.: C. W., Indo-European Origins ofthe Celtl:c Verb 1. The sigrnatic

aorist (1962). Watkins, Idg. Gr. (lII. 1): C. W., Indogermanische Grammatik lII. 1, Geschichte del'

indogermanischen Verbalflexion (1969). Wissmann, Jlfomina postverbalia: W. W., Nomina postverbalia in den altgermani­

schen Sprachen, 1. Teil (1932). ZCP,' Zeitschrift für celtische Philologie. Zinkevicius, Líet. Dial.: Z. Z., Liet7Wi1? Dialektol<Jgija (1966).

11

Page 6: Jasanoff.1978.Stative and Middle in Indo-European

r I ! ¡ I \

I I

1

INTRODUCTION:

THE STATIVE FORMATIONS OF INDO-EUROPEAN

§ 1. Among the athematic root presents reconstructible for Indo-Euro­pean are several which denote a state rather than an action or process. Such presents, which clearly belong to an ancient type, may be active or middle, cf. Ved. váo$ti 'wishes', Hitt. wekzi, Gk. ÉXWV ( < *~ek-); Ved. ko$éti 'dwells', GIL XTlfL€VOC; «*tkYei-); Ved. cáo$te 'sees', Av. Caste «*k¡'eks-); Ved. sáye 'lies', Hitt. kitta(ri) , Gk. X¡;;'i.'TI):¡. Forms of this kind survive as scattered archaisms in the attested lE languages, where their stative function is synchronically predictable only from the lexical meaning of the underlying root.

Even in such archaic groups as Indo-Iranian, Anatolian and Greek, however, unmotivated root presents account for only a small fraction of the stative verb forms actually found. In every early lE language the majority of statives are not primary, but morphologically derived: the starting point for the creation ofnew forms may be either a non-stative verbal root, as, e. g., in Ved. tro$yati 'is thirsty' < tro$- 'become thirsty, turn dry', Gk. ~aye: (perf.) 'is broken' < (f)l):y- 'break', Lith. turi 'has' < tver- 'seize', or a nominal stem, as, e. g., in Gk. ~1):(nAe:ú(Í) '1 am king' < ~I):O'¡A€ÚC;, Lat. albeiJ '1 am white' < albus, Go. -jJiwan, -aijJ 'be a servant' < jJiU8 'servant', etc. The formal devices used for this purpose in the daughter languages are extremely numerous. Most of the synchronically productive processes, especially in the later languages, are denominal in origin and perform a characteristically wide range of non-stative functions as well: compare, e. g., the twofold value of Gk. ~e:(¡)v(~w '1 am a stranger' and '1 receive as a stranger' < ~&voc;, xoup(~w '1 am young' and '1 bring up from boyhood' < xoüpoc;, xoupUhoc;, or OIr .. úraigedal' 'is green' < úr 'fresh, green', but . ailegedar 'alters' < aile 'other'. Denominatives like these characteristical­ly represent specializations of the lE denominative type in *-jejo- (cf. Ved. ukt¡anyá- 'act like a bull' <uko$án-, Hitt. nalJsariya- 'fear' < nalJlal', GIL Ep(~W < ep¡c;, etc.); while they have often had complex histories in the individual languages, their structure and status present no serious problems from an lE point.ofview.

13

Page 7: Jasanoff.1978.Stative and Middle in Indo-European

The morphological origins ofthe deverbative, 01' "primary" 1) statives of the historicallanguages, on the other hand, are considerably more obscure. With the exception of the descendants of the lE perfect, which are usually historically transparent, almost all the non-denominative stative forma­tions ofthe older lE traditions have at one time 01' another been linked to the supposed existence in Indo-European of a class of "e:-statives" 01' "e­verbs"2). The problems surrounding this category constitute the major focus ofthe present study, and will be surveyed in a preliminary fashion in §§5-13.

§ 2. The lE perfect (whence the perfect active ofGreek and Indo-Iranian) originally denoted the state resulting from the accomplishment of an action 01' process. This value is still faithfullypreserved in Homeric Greek, where it is evident in such forms as OAWAO( '1 am lost' < OAAU(.lo(L, 7tÉ7tOL!)o¡ '1 trust' < m:Wo(.lO(L, ~a'1:'f)){.O( '1 stand' < la't'O((.lO(L, 't'ÉIl'l'f)){.O( '1 am dead' < e'l~a){.w; the "resultative" perfect (e. g., 3É~w){.O( '1 have given', yÉypO(rpO( '1 have written', etc.) is almost wholly a post-Homeric development. The perfect is likewise well-attested as a stative in both branches of Indo-Iranian, although in Vedic, as in later Greek, its inherited function has in large part been replaced by that of a general preterite, cf. Ved. bibhaya 'is afraid' < bh'í-, dadhárt;a 'is bold' <dhrt;-, jagára 'is watchful' <gr-, but cakára 'did', jaghána 'killed' etc.; Av. tütauua V. 6. 32 'is possible' < taúu-, 3 pI. caxnarfi Y.44.13 'care about' < kan-, ahisaiiaY. 29.1 'oppresses' <ha(ii)-. Outside Greek and Indo-Iranian, the stative perfect has left substantial traces in Germanic, where preterito-presents such as Go. wait '1 know' ( = Gk. ol80(, Ved. védc(,) , man 'T intend' (= Gk. (.lÉ(.lO'lo(. Lat. memin'i) , ga-dars '1 dare' (= Ved. dadhárt¡a) constitute an archaic category of considerable descrip­tive importance. Elsewhere in the family only scattered forms continue the perfect in its primitive function. Sorne of these, like Lat. memint, (g)nOut, 15m, OCS vede '1 know' « *yoidai) and Hitt. saklJi 'id.' « *-ljai) are inherited perfects in shape, secondarily reinforced in their original value by the addition of the hic et nunc particle *i; others, like 011'. ad'ágathar 'fears', OLith.líekti 'is left', stóvi 'stands', OCS boit'b se 'fears' and Toch. B nesam 'is' ( < *nos-) are synchronically indistinguishabl~ from presents, but appe~r to continue inherited perfects in a morphologically altered formo

1) Rere and below, the term "primary stative" will be used to mean a stative verb derived directly from a verbal root, rather than from an already characterized (typically nominal) stem. .

2) It is, of course, potentially misleading to speakof "e-verbs" rather than "e­presents" 01' "e-aorists" in Indo-European; nevertheless, the term will be retained as a convenience in the discussion below.

14

r § 3. The lE perfect is formally characterized by o : ~ apophony (cf. Gk. oí~O( : 'L~fJ.E'I = Ved. véda : vidmá = Go. wait : witwrn) and by a distinctive set of personal endings; it is less clear whether reduplication, often absent in the daughter languages, was obligatory 01' merely facultative in Indo-Euro­pean. Our understanding of the position of the perfect in the system of the lE verb has been enormously furthered by the discovery, made indepen­dently by J. Kurylowicz, BSL33, 1---4 (1932) and Chr. S. Stang, NTS6, 29ff. (1932), that the perfect endings are etymologically related to those of the present and aorist middle. The language which supplies the decisive evidence in this regard is Hittite, where the simplest forms of the middle endings in the singular, 1 -(lJ)lja, 2 -(t)ta, 3 -a, differ only in the vocalism of the 3 sg. frQm the classically reconstructed perfect endings sg. 1 *-a « *hze), 2 *-tha ( < *-thze), 3 *_e 3

). Outside Hittite, the perfect and middle agree further in their common predilection for r-endings in the 3 pI., cf. Ved. 3 pI. pf. -ur, Lat. -ere; Ved. 3 pI. mido -re, -ra[n] , Av. -aire, Toch. B -re.

Predictably, the historical affinity of the perfect and middle endings is reflected in their respective distributions. In Greek and Indo-Iranian the perfect archaically behaves as part of the middle paradigm: Gk. OAWAO(

corresponds semantically not to the active OAAU(.lL but to the middle OAAU(.lo(L,

while deponents like y[Y'lO(.lO(L and (.ld'lO(.lo(L form "active" perfects yÉyo'lo( and (.lÉ(.l·I)'Io(, respectively; in Vedic the presents corresponding to rnamárt;a 'ignored', vavárta 'turned' and ruróca 'shone' are rnrt¡yate, vártate and j'ócate, all media tantum. An especially common lE pattern opposes a middle root aorist, denoting entry into a state, to a perfect, representing the state itself; characteristic pairs ofthiskind are *g~lS(t)Ó 'took aliking to' (cf. Ved. 3pI. ájut;ran): *(ge)góuse 'enjoys' (cf. Ved. jújot;a(ti)), *mt1(t)ó (*mén(t)o) 'brcíught to mind' (cf. Ved. ámata, GAv. 'manta) : *( me )móne 'is mindful (of)' (cf. Gk. (.lÉ(.lO'lE, Lat.meminit, Go. man), *li7c"(t)ó 'left (intr.)' (cf. Ved. 2sg. rikthálJ,): *(le)lóik1'e 'is gone' (cf. Gk. intr. W,OmE'I), *sthz(t)ó 'stood up' (cf. Ved. 3 pI. ásthiran, Toch. B ste 'is' < pre-Toch. *sto): *(ste)stóhzy(e) 'stands' (cf. Ved. tasthríu, Lith. stóvi).

§ 4. In the historical lE languages the perfect has characteristically come to acquire the value of a general preterite, thereby merging semantically with the aorist and/or imperfecto This function is already dominant in the language of the Rigveda, as well as in the oldest !talic, Celtic and Germanic; in Greek and lranian the evolution from stative to preterite falls within the historical periodo An identical development may perhaps be assumed in part for Anatolian, where, as recently argued by H. Eichner, Flexion und Wortbild~tn{J, 87ff., and E. Risch, ibid., 251ff., the non-stative value of lJ,i-conjugation forms like 1 sg. aklji '1 die' ( < *-lje < *-ljai) can be accounted for by assuming that original perfects like *ak7Ja ( < lE *-hze),

3) The form of the middle endings is discussed at greater length in §§ 45---47.

15

Page 8: Jasanoff.1978.Stative and Middle in Indo-European

n. '1 am dead', regularly became preterites ('1 (have) died'), from which ~ew presents were created by the addition of the hic et nunc particle *i.

The loss of the perfect in its original function has been offset in most languages by the growth ofnew stative categories oflate 01' "dialectal" lE date. A conspicuous example of such a formation is the perfect middle, a productive type in Indo-Iranian and Greek, and apparently represented in Old Irish as well (cf. § 68). Both the form and function of the perfect middle in these languages suggest a comparatively recent creation. From a historical point ofview the middle endings are redundant in the inflection ofthe perfect; as we have noted, the morphological and semantic ties ofthe perfect "active" are not properly with the active atall, but with the middle. In both Indo-Iranian and Greek the typical function ofthe perfect middle is to renew the lE perfect in its original meaning, while the inherited perfect itself tends increasingly to acquire new and specifically active 01'

non-stative functions. Verbal roots which form active presents in Vedic Sanskrit typically utilize the perfect active as a simple preterite, essentially equivalent to the imperfect; in such cases the perfect middle is often stative, cf. paprátha '(caused to) spread out' beside paprathé 'is stretched out', vavárdha 'made great' beside vavrdhé 'is great, has become great', jajana 'begat' (contrast Gk. yÉyove) beside jajñé 'has been born'. The situation in Greek is similar: although the stative value of the perfect active is still consistently maintained in Homer, the tendency of the perfect to acquire middle forms in its intransitive uses, particularly in the presence of a well-developed middle paradigm elsewhere, is already pronounced. Cf. TÉ'rUXTOCL beside TeTeuxw<;, dfLOCPTOCL beside b:fLfLope, l:ípwpe beside ÓpWp'l)TOCL (subj.), later 1tÉ1teLIJ"fLOCL (Aesch., etc.) beside 1tÉ1tOL6oc, etc. For the history of the perfect middle in Indo-Iranian and Greek and its distribution in relation to the active see further Renou, Val. du parj., chs. 5-8, and Chantraine, Hist. du parf. gr., chs. 3,5. The status ofthe perfect middle in Indo-European is discussed in §§ 71, 101.

§ 5. More conspicuous than the remains of the perfect in many lE languages are the stative formations which appear to contain a suffix *-e-. The problems surrounding the form and distribution of this mQrpheme rank among the most stubborn in lE comparative grammar. This is due partly to the fact that Indo-Iranian, normally our best source of information about the lE verbal system, offers no unambiguous traces of stative *-e- at all; more than tbis, however, it reflects the extraordinary degree to which the categories allegedly built on this suffix in the remaining languages differ among themselves. Before proceeding further, it will be useful to review the basic evidence for "e-verbs" in Indo-European. Both deverbative and denominative formations occur. The former, which are more difficult in almost every respect, will claim the major share of our attention in subsequent chapters; the latter, which present a more

16

...

coherent formal picture, provide a more convenient point of departure for the survey which follows.

§ 6. The lE denominative sta ti ves in *-e- have recently been studied by C. Watkins, TPS, 1971, 51-93, who has demonstrated that they are more widely distributed than has traditionally been supposed.

The clearest examples of denominative e-statives are found in Balto­Slavic and Italic. Balto-Slavic has a productive class of denominative verbs with infinitives in *-eti (Lith. -é%i, OCS -éti) and present stems in *-eje(o­(Lith. -éJa-, OCS -éje(o-); when not compounded with a preposition, these are typically stative in Slavic (cf. OCS bogatéti, 3 sg. -éjet'h 'be rich' < bogat'h 'rich'; ~lméti, 3 sg. -éjet'h 'understand, know how' < um'h 'understanding') but more often inchoative in Baltic (cf. Lith. senéti, 3p. -éJa 'grow older' < senas 'old' ;j~lOdéli, 3 p. -éJa 'turn black' <juodas 'black'). The same two functions are associatéd with the denominative suffix *-e- in Latin; here, however, statives in -ea, -ere (e. g., senere < senex, nigrere < niger, silere < *si-lo-; cf. Pokorny, IEW, 891) are formally distinguished from inchoatives in -esca, -escere (i. e., *-e-ske(o-; cf. senescere, nigrescere, silesce­re). Despite the frequency of attempts to derive both Lith. senéli and Lat. sene-re from the lE denominative type in thematic vowel + *-je(o- (cf. GIL (fnAÉw < cpLAO<;, Ved. asvayáti 'desires horses' < áSva-, etc.), Watkins has shown that the only plausible source ofthe Balto-Slavic and Latin forms is an inherited present class in *-e- 01' *-e-je(o-. A specifically athematic paradigm is suggested by the Latin 3 pI. ending -ent ( < *-enti), beside which 1 sg. -eaand the thematic inflection ofthe corresponding Balto-Slavic forms are likely to represent innovations.

A major finding ofWatkins' study is that Hittite dénominative statives such as 3 sg. nakkezi 'is important' < nakki- 'important', tannattezzi 'is deserted' < tannatta- 'deserted', lJassuet 'was king' < lJass~l- 'king' are reflexes of the same historical category. It is significant that these and similar forms are found side by side with inchoatives in -e.s- (i. e., *-e:..s-): nakkeszi 'becomes important' and tannatte.szi 'becomes deserted' agree completely in function with the Latin type in -escO, and stand to it in the same formal relationship as, e. g., Hitt. pa7Js- 'protect' to La,t. paseO. The extended suffix-form *-e-s- is, in Watkins' view, also to be seen in Greek sigmatic aorists snch as 6OCp~~crOCL, pLY'~crOCL, O:V6'~crOCL, etc.; the corresponding presents 6ocp~Éw, pLyÉW, o:v6Éw, etc., although usnally taken as denominatives to nenter s-stems, are for the most part chronologically later than their associated aorists, and appear to have been created secondarily on the model of pairs like cpLMw : cpLA·~crOCL. A clear case of a denominative e:..stative in Celtic is 011' .. ruidi 'blushes' (AH), which invites direct comparison with Lat. t'1lbere and Russ. Ch. SI. ndéti s~, 3 sg. -éjet'h 8~ 'id.'; nnfortunately, the complex origin of the Celtic weak verbs in -i- makes it difficnlt to distinguish further examples of this type from original presents in *-eje(o­and *-ije(o-. The Germanic clenominative verbs of the third weak class

17

Page 9: Jasanoff.1978.Stative and Middle in Indo-European

..-------------~r .. ------------(type OHG roten, 3 sg. rotet ( < *-aip) 'turns red') testify to the original presence of the same category in a sixth branch of the family as well; here, however, the formal relationship between the predesinential diphthong *-ai- and the lE suffix *-e- has never been satisfactorily explained (see below). Whether denominative statives in *-e- are also preserved in the Armenian present class in -i- (type unim '1 have') is unclear; on the class as a whole see now Godel, Introd. to the Study oj Class. Arm., 155.

§ 7. The synchronically deverbative 01' "primary" formations in *-e­present a more varied picture. In Latin the deverbative statives of the second conjugation,-such as habere, manere, tacere, etc., are indistinguish­able in the present system from denominatives like senere, rubere, etc., and like these, appear to continue an athematic type in *-e-. A similar type has been identified in Tocharian, where Schmid, Stud. z. balto ~L idg. Vb., 99f. and Watkins, Celt. Vb., 70f. have derived the stem vowel of class III presents (type 3 sg. A b'ikatar, B triketar 'is confused') from lE *-e-; the absence ofroot-final palatalization in theseforms, however, makes such an interpretation doubtful (cf. §20).

N one of the other branches of lndo-European offer clear cases of a non­denominative present class in *-e-

q'). In Greek it is usual to take lE *-e- as

the source of the aorist passive in -'f)-: the originally medial rather than passive function of this category is evident from forms like [J.tyr¡ 'mingled', E<X"(f) 'broke', <.páv'f) 'appeared', etc. (cf. Schwyzer, Gr. Gram., 756ff.; Chantraille, G"i'am. hom., 399ff.). Such\orists are found in association with a variety of different present formations (cf. [J.tcrye;·w;¡, IXYVUTtX¡, <.pdve;TtX¡); especially well-represented among these is the intransitive type in *-iejo-, seen not only in <.pdVE;TtX¡: <.páv'f), but also in [J.dVE:TtX¡: E[J.áv'f), xtXtpe;¡: XáP'f), etc.

The situation in Greek recalls that in Balto-Slavic, where *-e- in primary statives is confined to the infinitive stem, historically derived from the aorist (cf. Lith. minéli 'think, remember', pret. minéjo; OCS mb1u!ti 'id.', aor. mbne). Unlike Greek, however, Balto-Rlavic opposes its extra-present forms in *-e- to presents of a uniform type, which are characterized by *-i- in Baltic (cf. Lith. 1 sg. miniú, 3p. mini) and by *-i in Slavic (cf. OCS 1 sg. mbñjq, 3 sg. 'fnbnit'b). These forms are not immediately reconcilable with each other and neither Baltic *-i- nor Slavic *-i'- has a clear source in llldo­European. 'In this respect, if in no other, the ejt-verbs ofBalto~Slavic invite comparison with their functional counterparts in Germanic. Here, as in Latin, no distinction is made between the deverbative and denominative types: like roten, OHG haben 'have' inflects as a verb ofthe third weak class, showing -e- < *-(¿i- before the present endings and the suffix of the dental preterite (habeta, roteta). The origin of this etymological diphthong, as remarked aboye, is disputed, and the historical position of the third class is further obscured by the fact that outside Old High German *-ni-

4) On the supposed reflexes of such forros in Ce1tic see ch. 4, fn. 10.

18

participates in an exceptional alternation with *-n- in Gothic and Scandinavian (cf. Go. lsg. hnbn, 3 pI. hnband, 01 1 sg. hej, 3 pI. hnja, etc.) and *-jn- in Old Saxon and Old English (cf. OS 1 sg. hebbi'u, pI. hebbind, OE 1 sg. hcebbe, pI. hnbbap, etc.). Although a direct relationship between the Germanic forms and the e-formations surveyed aboye is suggested by their close semantic agreement, it should be noted that nowhere in the Germanic verbal system is an unequivocal reflex of lE *-e- preserved 5).

§ 8. Thus, while the denominative statives discussed in § 6 seem ultimate­ly to poin t to an lE present type in *-e-, the only branch of the family which preserves unambiguous traces of a corresponding class of deverbative presents in -*-e- lS Italic. Greek and Balto-Slavic, by contrast, appear at least synchronically to exclude *-e:. from the present system of non­denominative statives; whether this 01' the Italic pattern represents the more ancient situation has never been conclusively resolved. According to one prevalent conception, the "e-verbs" appear in their most archaic form in Greek, where -r¡- is confined to the aorist and plays no role in the formation of the associated parent stem; advocates of this position have typically attempted to explain the Balto-Slavic t-presents, and occasional­ly the Germanic statives as well, without recourse to the assumption of inherited present forms in *-e-. Proponents ofthe opposite view, taking the exclusion of *-e- fram the present in Greek to be secondary, have instead sought to derive the Balto-Slavic and Germanic forms from prototypes ultimately equatable with Lat. hab"i're, manere, etc. Numerous variations of these two basic approaches, embracing a wide range ofindividual analyses, have been proposed ayer the past century; none, however, has succeeded in winning general acceptance.

Since the stative presents ofBalto-Slavic and Germanic clearly hold the key to a proper understanding of the position of the "e-verbs" in lndo­European, it will be useful here to give a brief survey of earlier attempts to explain these formations. Our purpose in so doing will not be to achieve bibliographic or factual completeness, but simply to focus attention on a group ofproblems which will be addressed at greater length in the chapters below.

§ 9. Among the historical explanations which have been proposed for the Balto-Slavic stative presents in *-t-, one of the most widely accepted identifies these forms with the lE type seen, e. g., in Ved. mányate 'thinks', Gk. [J.tXtVE;TtX¡ and 011'. do' moinethar 'remembers'. Presents of this class, characterized by zero-grade of the root and a suffix traditionally recon­structed *-jejo-, overlap closely in function with the stative formations just

5) To be sure, *-e- has 1eft a clear trace in Go,fahejJs 'joy', but this is a deverbative llOUll. It is discussed in § 56,

Page 10: Jasanoff.1978.Stative and Middle in Indo-European

discussed, as can be seen aboye all from such correspondences as Ved. lúbhya- 'be desir??s': Lat. lubet, .Go. *lu?ct:ij>. (cf. ~~lbai~~ '~w~~'),;. Ved. búdhya-, Av. büiona- 'be awake': LIth. budeh, -~, OCS b'bdet~, -~t'b Id. ,Ved. grdhya- 'be greedy': Serb. Ch. SI. z1:Idéti, -it'b 'des~re'.; Ved. tz't;yati 'be thirsty', Go. j>aursjan silc 'id.': Lat. torrere, e~c.6). SImIlar forms, though most numerous in Indo-Iranian, areattested m every early lE language; like the semantically related perfect, presents ofthe mányate: fLdve'l"o(L type are often found in association with deponent root aorists, as can be seen from such,cases as Ved. aor. 3 pI. ábudhmn, ptcp. budhaná- beside búdhya-, aor. ptcp. trt;arJá- beside trt;ya-, aor. 3 sg. ámata, ptcp. mananá-, GAv. manta beside mányate, etc. .

An exact equation of Lith. 3 p. mini, OCS 3 sg. mhnÜ'b with forms hke Ved. mányate is ruled out by the fact that the historically regular treatm~nt of lE *-jefo- in Balto-Slavic is *-j~fa- afte~ ;.?,;el~ ~nd *-~~:~, wIth palatalization after consonants, cf. Llth. 1 sg. l~ez~u 1 hck ,3 p. hezw, OCS lizQ 'id.' 3 sg. Úzet'b. Since the first edition ofBrugmann's Gnlndr~f3 (1892), however, an important group of scholars have held .that both the Bal~o­Slavic ~-presents and the mányate: fLdve'l"o(L type contmue an lE formatlOn in which the suffix-form *-jo- (*ijo-) regularly alternated not with *-je­(*-iie-) but with *-t-, under conditions parallel to the alternation of *-0-and *-~- elsewhere. Such "semithematic" presents, in Brugmann's view (Gr. 1 22

, 1055-7), were originally distinct from the fuIly the~atic, non­statíve type in *-jefo-, with which they generally feIl together m every lE dialect except Balto-Slavic. . . . ..

Although embraced by such distingmshed authorltws ~s A~tome Mmllet (see, e. g., his Slav. com., 232-4) anel Chr. S. Stang (e. g., m hlS Vergl: G1·. d. balto Spr., 319-20), the semithematic theory has .ah:ays r.emame.d a minority view. The alleged evidence for semithematlC mflectlOn o.f *e!o­verbs outside Balto-Slavic (cf. Lat. capio, -is, a,udio, -'is; OIr. 'ga~b~~l '1 take', 3 sg. 'gaib, 'léiciu '1 leave', 3 sg. 'léici; OHG liggtl '1 líe'.' 3 sg. ligit, Go. hausja '1 hear', 3 sg. h(l~lseij» is at best inconclusive: nothmg prevents a deri~ation of the stem vowel *-~- in Italic, Celtic and Germanic from lE *-ie- or its Sievers' Law variant *-ije-7

). Only fully thematic equivalents of these forms are found in Indo-Iranian, Greek and Hittite, and none of these languages preserves any traces of a s~m~thematic co~jugational patte~'ll elsewhere in its verbal system. Even wlthm Balto-SlavlC, as we shal~ see m ch. 4, ~-presents fail to show the predicted distribut:on o~ the~a~~c and athematic forms, the former being found in the 1 sg. only (Llth. m'¿mu" OCS mb1ijQ; cf. § 82). Thus, the semithematic theory is essentiaIly an arbItrary

6) Further sueh pairs are lis~ed by Sehm~d, St1.td. Z. balto .1(, idg. Vb., eh.y .. 7) 01', b Celtie and Germame, from the lteratlV~-eausatlve and,,~enOlmna~!:e

suffix *-ejejo-, whieh in faet aeeounts for the largest smglenumber of ~-presents m these languages.

20

r compromise, motivated only by a perceived need to reconcile the athematic appearance ofthe Balto-Slavic statives with the thematic inflection ofjefo-presents in the remaining languages.

§ 10. Among other attempts to explain the ~-presents \vithout reference to lE *-e-, it will suffice here to mentíon only two. In Infl. Cat., 79-84, Kurylowicz has sought to derive the theme vowel of the Baltic and Slavic forms from the 3 sg. perfect ending *-ei (i. e., *-e + hic et nunc *i). Since final *-ei would in fact regularly yield -i in Lithuanian and -i(t'b) in Old Church Slavonic, the quantitative difference between, e. g., Lith. gaTi 'bums' (-i) and OCS g01'it'b (*-'i-) would have an attractively simple explanation under this theory. A se,rious difficulty, however, is that the Lithuanian reduction of *-ei to -i 'in final position is not a development of Common Baltic date; consequently, it is questionable whether the 3 sg. perfect can have played a role in the creation of the obviously related i-presents of dialectal Latvian and Old Prussian (cf. Latv. gulim 'we lie', OPr. t~lTl'imai 'wir sollen', and see also § 85). Less daring, though hardly more compelling, is the view of F. Specht, whose major study of the e-statives (KZ 62, 29-115 [1935]) a.ttempts to show that stative *-e- was originally confined to extra-present ((l~lj3eTpl'tisentisch) functions, while its place in the present system was taken by the thematic suffix *-ejefo-. Specht views *-eje- as the direct source of Slavic -i-, comparing the apparently parallel development of *-eje- to *-'i- in inherited iterative-causatives like OCS nosit'b 'canies' < *nokeje(ti). From a phonological point ofview, however, the contraction of *-eje- to *-'i- cannot be regarded as firmly established, while the presumed subsequent shortening of *-'i- to -i- in Baltic seems even less secure (see §§ 83,86). Nor is it clear to what extent *-ejefo- can be assigned stative value in Indo-European : we have seen that denominative presents like Lat. albere are better regarded as continuing preforms in *-ejefo- (§6).

§ 11. Attempts to account for the class III weak verbs of Germanio along similar lines have not been numerous. The present writer's derivation of Gmc. *-aij> from a 3 sg. middle in *-ai ( < lE *-oi), secondarily suffixed by the active 3 sg. ending *-j> ( < *-ti) (Lg.49, 850-70 [1973]), is discussed in ch.3. Of the theories summarized aboye, only Specht's is applicable to Germanic: hereforms like 1 sg. OS hebbúl, OE hillbbe (§ 7) arephonologically derivable from *kapéjo, while Gmc. 3 sg. *habaij> can be taken to reflect the substitution of *kap¿jeti for *ka,péjeti undel' the influence of the non­present stem in *-e:- (Specht, op. cit., 77). This, however, fails to motivate the absence of -j- in Go. 1 sg. haba < *ha,b(J and leaves the reasons for the retention of *-e- in part of the paradigm and its replacement by *-e­elsewhere unexplained. Given the poor evidence for statives in *-ejefo­outside Germanic in any case, Specht's views have not been widely accepted.

The great majority of scholars who have investigated the third weak class have instead taken the position that *-e-, 01' a v<j,riant of this suffix (see

21

Page 11: Jasanoff.1978.Stative and Middle in Indo-European

below), was originally distinctive of the entire category. The standard handbooks of Germanic, from Streitberg's Urgermanische Gmmmatik (1st ed. 1896) to Guxman's Smvnitel'naja gmmmatika germanskix jazykov (1966), are virtually unanimous in deriving the Germanic forms in *-ai­from lE *-eje-, although the supposition that such a sequence would have yielded a Germanic diphthong is at least questionable (cf. § 52). Even under a derivation of *habaip < *kap~jeti, however, the position offorms such as Go. 1sg. haba, 3pl. haband and OS hebbin, hebbiad remains unexplained. According to a widespread older view, still favored by Krahe, Germ. Spmchw. II, 121, Go. haband regularly continues an athematic 3pl. *lcap~nti, parallel to Lat. habent; the agreement of 3 pl.-cmd < *-enti with -and < *-onti would then have led to the secondary creation of quasi­thematic forms 1sg. haba and 1pl. habam. Yet it is highly doubtful that *-ent- would have yielded *-and- and not *-ind- ( < *-ent-) in Germanic, and equally unclear how 3 sg. *kap~jeti and 3 pI. *lcap~nti could ever have come to be associated in the same paradigm 8). A number of recent studies have therefore uplield the historical priority of forms like OS hebbin, ayer Go. haba, although they have rejected the specific interpretation placed on hebbiu by Specht. Characteristically, such theories have instead assumed that the ja-forms of Germanic contain an apophonic variant of the stative suffix itself; as might be expected, this assumption has had important consequences for Balto-Slavic as well.

§ 12. The suggestion thi1t the stative suffix originally displayed quantita­tive apophony in Indo-European is hardly new. Such a position was taken, e. g., by Brugmann in the second edition ofthe Grnndrij3, 23

, 195ff., where the e-statives of the attested languages were derived from "disyllabic heavy bases" in *-e1:; the reduced-grade (*-i < *-'i1i) and zero-grade (*-i) of this formant were, in Brugmann's opinion, most clearly recognizable in the semithematic presents of Baltic and Slavic, but also detectable in OS hebbiu, etc. Although the apophonic theory on which this interpretation is based is no longer tenable, it. has had several modern representatives. H. Wagner, comparing the Hittite type 3 sg.lJ,alzai 'calls', 3 pI. lJ,alziyanzi (e­Verba, 50-2) reconstructs a pre-Germanic present in *-eje-j*-jo-, which he supposes to have been secondarily thematized from an earlier type in *-ei-j *-i-. A similar position is taken by W. P. Schmid, who assumes an athematicpresentin 3 sg. *-~iti, 3pl. *-'i1jénti (Stnd. z. balto n. idg. Vb., 83); he then derives Slavic *-i- and Baltic -i- from the suffix-forms *-ei- and *-'i1i~, respectively, and sees a thematized form ofthe latter alternant as the basis ofthe Indo-Iranian type in -ya- (mányate). J. Puhvel, Lar. and the lE Vb., 53ff., sets up an inherited type in *-eEY-j*-EY-, the "palatal" laryngeal of

8) For other attempts to explain the inflection of the third weak class on the basis of an athematic type in *-e- see § 52.

22

which would allegedly yield -í- in Baltic; how Puhvel would explain Slavic *-i- and Germa.nic *-ai-j*-(j)a- is not made clear.

Such theories, even if they were free of phonological difficulties, would hardly be more than stopgap measures. In order to maintain the existen ce of athematic presents in *-ei- (*-eEY-) j *-'i1i- (*-t-, *-EY-) in lndo-European, it is necessary to as sume their complete elimination as a distinctive type in Indo-Iranian, Greek and, pace Wagner (cf. §54), Anatolian-precisely the three branches of Indo-European in which the athematic conjugation is most faithfully preserved. The suffix *-ei- assumed by Wagner and Schmid is nowhere unambiguously attested as such, and in its long diphthong differs radically from every other affix in the lE verbal system; a reconstruction such as Puhvel's escapes the latter difficulties, but only at the cost of assuming an ad hoc and very doubtful phonological rule~

In some respects the most attractive of the "ablaut" theories of the e­statives is that put forth by W. Cowgill, Lg. 39,265--6 (1963). Basing his views in part on a suggestion ofW. Bennett (cf. §55), Cowgill proposes to derive both the t-presents of Balto-Slavic and the third weak class in Germanic from an earlier thematic type in *-'i1jejo-, where *-'i1- represents the zero-grade of *-e- (i. e., *-ehr ). This analysis has the merit of simplicity at the morphological level: if the phonological developments which it presupposes could be shown to be correct, the stative presents of the northern lE languages would represent jejo-presents of a type routinely encountered throughout the family. The developments in question, however . are by no means trivial, and appear to be contradicted by forms in which a vocalic laryngeal is lost without a trace befare *-jejo- in both Balto­Slavic and Germanic. Such cases, and Cowgill's theory generally, will best be deferred to chs. 3 und 4, where they will be discussed in detaiI.

§ 13. The e-statives of Balto-Slavic and Germanic thus remain in many key points obscure. Closely associated with their elusive history are a host of related questions, bearing on such unsettled issues as 1) the inherited distribution ofthe stative suffix *-e-, 2) the original form ofthis suffix and its relationship to the morpheme traditionally reconstructed *-jejo-, and 3) the ancient position of the e-statives vis-a-vis their close functional counterpart, the perfecto

This study will attempt to resolve some of these uncertainties by proposing a new explanation for the Germanic and Balto-Slavic statives: more specifically, an effort will by made to show that a close analogue to these formations exists as a productive category in Indo-lranian and Greek. Rather than proceed directly to the problematic forms themselves, however, we shall turn first to two lE languages which have never played a major role in discussions of the e-verbs, but which present the comparatist with a verbal class ofunsuspected interest for their study. These languages are Tocharian A and B; the forms in question are those of Krause's third present class, a detailed discussion of which follows.

23

Page 12: Jasanoff.1978.Stative and Middle in Indo-European

II

CLASS III PRESENTS IN TOCHARIAN

§ 14. The Tocharian presents of W. Krause's third' present class (cf. WToch. Gram. I, 66~8 for Toch. E; for Toch. A and E cf. Krause~Thomas, Toch. Elem. I, 200~1) have two defining formal characteristics. Common to the entire class is a stem~final element which appears as ~a~ in Toch. A and ~e~ in Toch. E, pointing to a Common Tocharian vowel which, in accordance with a suggestion of J. Schindler (personal communication), I shall represent by the symbol *ií. Equally distinctive is the virtually obligatory presence of the middle endings; the rare instances of class III forms with active inflection constitute a case apart and will be discussed separately in § 39. The class III paradigm may be illustrated with the present forms of the common verb AE mask~ 'be, be located, sich befinden' 1) :

A E sg.l maskamar maskemar

2 maskatéir masketar 3 maskatar masketar

pI. 1 maskamtar maskemt(t)ar 2 maskacar masketar 3 maskantCir maskentar

ptcp. maskama'Y[!, maskemane ger. I maskal maskelle inf. maskatsi [maskatsi] 2)

Such presents correspond to class V (~a~) subjunctives and class I preterites, both of which may be active as well as middle; the only exception is the root mask~ itself, which anomalously forms a class III preterite in Toch. A (3 sg. maskas beside E maska).

1) Here and below, forms which ~re not directly attested but which can be predicted with complete confidence from other lexical items are not marked with an asterisk.

2) Formed, as regulatly in Toch. B, on the subjunctive stem.

24

§ 15. The verbs for which class III presents are attested have been collected by Krause, loe. cit., for Toch. E, but no comparable inventory has been compiled for Toch. A. Eoth languages are taken into accoUllt in the list which follows.

Common to A and E: lcul~ 'subside'; kulyp~ 'desire'; trik~ ego astray, be confused'; triw~'mingle'; mask~ 'be, be located'; A niitsw~, E miitsts~ 'starve'; A ritw~, E ritt~ 'join with'; wilc~ 'vanish'; siitk~'spread out (intr.)'; tsar~ 'separate (intr.)'.

Found only in A: paz~ ego out, be extinguished'; park~ ego up'; mal~ 'become oppressed' (E class X); yu~ 'bend toward'; yutk~ 'worry'; lit~ 'fall down' (E lait~, IV); watk~ 'separate (intr.), decide' (E IX); wap~ 'weave'; .41t1·~ 'sorrow'; 8'¡k~ 'overflovv'.

Found only in E: kurp~ 'care about'; kramp~ 'become bothered'; tas~ 'resemble' (A téislc~, II)3); nu~ croar' (ñewetar; A VIII); palk~ 'bum (intr.)'; pra1Ík~ 'restrain oneself' (A I) ; prutk~ 'be fulfilled' ; pza1Ík~ 'be put on sale' ; ma1Ík~ 'be inferior' ; mars~ 'forget' (A VI); mit~ 'set out'; m1lsk~ 'disappear'; lip~ 'remain over'; lu~ 'send' (lyewetar); luk~ 'shine forth' (lY1lkettir); lyu~ 'rub'; waks~ 'be upset (?)'; spant~ 'trust'4); spark~ 'perish'; sru,k~ 'die'; tsa1ik~ 'arise' (tse1iketCir); tsam~'grow' (A IV); tsalp~ 'cross over, beredeemed' (AIV); tsu~ 'accommodate oneself'.

Here too may belong severa] presents which are too sparsely attested for an unambiguous determination of their class to be possible. E 3 pI. kwre'Y[!,ntar < kwar~ 'grow old' (A kul'~) and 1 pI. cukemar < tulc~ 'be hidden' (A tpulc~) could in principIe belong either to class II (thematic) 01' class III, although the zero~grade root vocalism of both forms makes class III the likelier choice5

). The class I subjunctive and class III preterite ofE pzatk~ 'come forth', on the other hand', speak in favor ofKrause's assignment of the participle E plyetkemane to class II. Special problems are presented by the anomalous behavior of A km'y~, E lce1'y~ 'laugh', which shows only active finite forms in both languages and appears to contract ~ya~ to ~e~ in A ptcp. karema'Y[!, (E keriyemane); see further § 39.

§ 16'. Formally indistinguishable from the class III presents are the class III subjunctives, which, like their indicative counterparts, are inflected by adding the middle endings to a stem ending in CToch. *~ií~. (The "active" class III subjunctives of Tocharian E (Krause~Thomas, 225), e. g., 3 sg. naka'Y[!" 1 pI. nlcem<naJc~ 'destroy', are athematic in form and properly belong to class I rather than here.) Such subjunctives are invariably intransitive; the majority correspond to intransitive and medial presents

3) Doubtful, see § 18. 4) Cf. A3pl. smiintantr~ii!rr¿, emended by Krause-Thomas (op. cit. II, 157) to

8piintatr~iif!1,.

5) For the initial palatalization compare B Zyuketiir < luk~.

25

Page 13: Jasanoff.1978.Stative and Middle in Indo-European

----------------------... ,.. ................ ....... of classes VIII .(-8-) and X (-nii8(k)-). Subjunctives of class III are reliably atteste.~fr~m eIght,roots: kiin- 'happen' (AB), kéi8- 'go out, be extinguished' (A~), tam- be ~or~ (A ~ sg. cmatiir, B C1netii1') , niik- 'destroy' (act.) / 'perish' (~m~.) (~B), nam- b.?V; .(~ only; A class V), piik- 'make ripe' (act.) / 'grow rlp~ (mld.) (AB), wal- dIe (A only), t8iik- 'burn (tr.)' (act.) / 'burn (intr.)' (mld.) (AB). Amnthroot, t8ii1'- 'separate', has in Toch. A agerundive IIsral ando a ve.rbal ~bs~~act sralune, which would normalIy imply a class III subJ:,tnctlve *8ratCtr. The 'presen~ of this verb, however, is t8ratiir (= B t8retar).' also of class IU; If genume, *sratliT would constitute the unique ('Xc~ptlO:r: to ~he general rule that class UI presents form class V subJu~ctIVes. SInce a class V subjunctive of t81iT- is in fact attested in B gel'. U t8ralle, the status of A sral, sralune must be considered doubtfuI.

§ 17. The presents assigned by Krause(-Thomas) to class IV (cf. WTock G1·. I, 68-9, Toch. Elem. I, 201-2) are historicaIly mere variants ofthose of class ~II. Like the forms surveyed aboye, class IV presents are almost excluslvely depon~nt; they are regula,rly accompanied by subjunctives of class Van? pretentes of clas~ I (a8- 'turn dry', with a class UI preterite in Toch. A, IS the only exceptlOn). The formal difference between the two types appears to be phonologicalIy conditioned : roots which form class IV presents .almost inv.ariably contain the underlying vowel a 01' the diph­~hongs a~ 01' au, whIle class UI presents typically show ii-, i- 01' u-vocal­Ism of the root syIlable.

. In Toch. B, where classes III and IV are more sharply differentiated than m Toch. A, the most characteristic feature of class IV is the stem-final voweI o (vs. class UI e), before which an a of the preceding sylIable is in sync~ll'onic terms, it~elfrounded to O. In Toch. A the class IV stem-vow~l is C:, as m class III; unhke the a of class IU, however, the cIass IV theme vowel IS r~gularly syn~opated ~efore an a in the next sylIable (i. e., before the . endmgs 1 sg. -mar, 2 sg. -ta1: and ~he participial suffix -miir(l,). Corresponding ~o the apparent umlaut of a to o m Toch. B, the root voweI ais repIaced by a In Toch.A.

The ~lassIV paradigm may be illustrated with tlw forms of plant- 'be pIeased : A

B sg. 1 plantmar plontomar

2 plantta1' plontotctr .3 plantatiir plontoUir

pI. 1 plantamtiir plontomt(t)iir 2 plantaciir plontoUir 3 plantantiir plontontiir

ptcp. plantma'f(i plontomane ger.I plantal plontolle inf. plantat8i [plantat8i]

26

I ! § 18. Class IV presents are attested from more than twenty roots, which

are Iisted below. Common to A and B: ar- 'stop'; a1't- 'praise'; a8- 'become dry'; A kla(w)-,

B klay- 'fall'; klaw- 'proclaim' ; A trap-, B t1'app- 'stumble'; A pot- B paut­'flatter'; plant- 'be pleased'; yat- 'be capable'; A 8partw-, B 8partt- 'turn'.

Found only in A: piir8k- 'fear' (pra8katii1'; B V); mlu8k- 'escape' (mlo8ka­tiir); wiiñk- 'gossip'; 8ak- 'remain'; t8arw- 'be confident'; t8iim- 'grow' (3 pI. Sctmanttir; BIII); t8iilp- 'cross over, be redeemed' (salpatliT; BIII).

Found only in B: aiw- 'devote oneself' ; karp- 'step down' (A VI) ; klaik8-'shrivel' ; klautk- 'turn' (A lotk-, VI); yank- 'delude'; lait- 'fall' (A lit-, III); wak- 'split, open (intr.)'; samp- 'be arrogant'; 8p1iw- 'run dry (? )'.

W orthy of note are several cases in which a class IV present in Toch. A corresponds to a class III present in Toch. B, and conversely. The contrast between B t8metii1', t8¿ilpetiir ( < t8iim-, t8iilp-; III) and A *samatiir, salpatii1' (IV) is clearly due to an original difference of apophony in the root syIlable: the Toch. B forms show the regular treatment of an earlier zero-grade, while the vocalism and initial palatalization ofToch. A sam-, salp- point to lengthened grade preforms *t8em-, *t8elp-. A similar relationship holds between A 3 pI. litantiir (III) and B 3 pI. laitontiir (IV), which appear to continue apophonic variants *lit- and *loit-. It is not impossible that, despite their semantic divergence, A *ywatiir (UI) < y~t- 'bend forward' and B aiwotiir (IV) < aiw- 'devote oneself' constitute yet another example of such a pair (cf. §38).

According to Krause--Thomas' classification, the class III.present B *ta8etiir (emended from ta8aitiir) has a class IV counterpart In A ptcp. ta8kmam. This, however, is almost surely not correct, since the regular class IV participle of ta8k- in Toch. A would have been *ta8kma:r" (cf. plantmiiíl'lJ). The unaltered root-vocalism of ta8kma'f(i clearly marks It as a class U (thematic) form - a fact which lends independent support to K. Schmidt's analysis (Flex1:on und Wortbildung, 287ff.) ofB ta8aitiir as a class U 3 du. 6

). .

§ 19. AH but a handful of the attested class III and IV ~reseI~ts ~re intransitive' while most denote an action 01' process, a substantIal mmonty are stative. The functional resemblance of these forms to the e-statives of Latin Germanic and BaIto-Slavic has been independently noted by Watkins, eelt. Vb., 70f., and Schmid, Stud. Z. balto u. idg. Vb., 99-100, bot.h of whom have attempted to explain the class III theme vowel on the basIs of an lE athematic type in *_e_ 7

). This suggestion is attractive in several

6) We could then assume that the root was originally task- in both languages, and that B tas- was extracted from forms like 3 sg. *taljtar < Pre-Toch. *taljljiitiir. According to H. Craig Melchert (personal co~munication), tiisk- is an sk-extension of the root which underlies Hitt. dakk- 'resemble'.

7) Watkins, TPS 1971, 61, hás retracted this view.

27

Page 14: Jasanoff.1978.Stative and Middle in Indo-European

r------

respects. A significant number of class nI (IV) presents are actually paralleled ?y apparent e:-forms elsewhere in the family; the examples quotable, mght m all, are AB as- 'become dry' : Lat. arere, AB kul- 'subside, slacken': Lith. guleli 'lie' (§ 38), A park- 'go up': Hitt. parkeszi 'becomes high', B palk- 'burn': Lat. fulgere, B mars-'forget': Hitt. mar'sezzi 'is false', B lip- 'remain over': Go. liban, -aip 'live' (probably also OCS pr'i-lhpéti 'stick;'), B luk- 'shine': Lat. [-ücer'e, B walc- 'split': Gk. (f)C(r~Vc(L It seems certam, moreover, that lE *eregularly yields Toch. A a and Toch. B e, cf. A mañ, B meñe 'moon' < *men-, A want, B yente 'wind' < *];tento-. Earlier attempts to identify the lE e:- statives in other Tocharian verbal categories, such as the class I (athematic) presents (H. Pedersen, Toch. 2, 162) 01' the class V (-a-) presents (e. g., J. Elfenbein, Ricerche Linguistiche 2, 185 [1951]), have not been successfuI. -

§ 20. ~ et des.pite.its ob.vious appeal, the Watkins----Schmid theory is open to a senous obJectlOn. Llke other etymological front vowels, the Tocharian reflex of lE *eregularly induces palatalization of a preceding consonant: compaiatively secure examples, beside A want, B yente just cited, include the nominative singular ofkinship terms like A pacar, B pacer 'father' (cf. A macar, B macer 'mother', etc.), the final syllable of which presupposes lE *-ter, and lengthened-grade presents like A *samatar, salpatar (see aboye) and B ñewetar 'roars', lyewetéi1' 'sends' (cf. §38). Pa.Iatalization is conspi­cuously missing, however, before the stem-vowel of classes In and IV, although it is theoretically possible that it was lost here by analogy, th~ complete absence of relic forms with a palatalized root-final consonant suggests rather ~h~t t~~ lack ofpa.latalization in these classes is phonologi­cally re?ul~r. I ~ lS SlWllfl~ant tha t m presents of the thematic type (class II) paIatalr.~a:lOn l~ ordlparrly preserved befare *a > *e (cf. 3 sg. A klyo~t¿¡r, B ~lyau~tar hears < *kle7tSet01'), even though unpalatalized forms are found m the 1 sg., 1 p~ and 3 pI. of the same paradigm (cf. 3 pI. A klyosantaT, B klyausentar' < *kleusont01') 8).

It is unlikely, therefore, that the class In (IV) theme vowel can be traced directly to lE *-e-. Furthermore, most ofthe hypothetical alternatives to *-e:- surveyed in the last chapter can be excluded for the same reason : such sequences as *-ei-, *-€le/o-, *-~-/*-jo- and *-eje/o-(cf. §§ 9-12) would, hke *-.e:- its~lf;almost s~rely have ca.used palatalization of a preceding consonant. Likewlse phonologl?ally ~nsatlsfactor~, although on different grounds, are the suffIX-forms *-,n-, poslted by SChlllId as the source ofthe Baltic statives in -i-, and *-'J-je/o-, taken by Cowgill as the prototype of both the Balto­Slavic presents in *-f- and the Germanic type in *-ai-/*-(j)a-. lE *-'Ji- would

8) Throughout this chapter we shall assume, at least as a convenient fiction, that the middle endings AB -tiir, -ntiir continue TE -tal', -nto1'. Nee fUl'ther ~4-(i. with note 51.

28

presumably have yielded CToch. *-i-, 01', with vocalization o~ the initial element, *-ai- « A *-e-, B *-ai-). The compound suffix *-'J-le/o- would probably first have given *-aja/a-, with the regular development of *'J to CToch. *a, Although no absolutely certain other examples of this sequence are known 9), it is attractive to regard certain class V presents, such as A r-wa- B 1'uwa- 'rip out' (: Lat. 1"lW, -er-e 'rip up') and AB wa- 'eat' (: OE ceow~n, Eng. chew), as original iteratives in *-aje/o- in which *-aja/a- has contracted to *-a-. Thus, the nature ofthe relationship, if any, between ~he class In (IV) presents and the e-statives of other lE languages remams very unclear.

§ 21. The q,bsence ofpalatalization befor~ the c¡ass IIIvt~eme vowelma~es it aH but certain that this vowel, representmg CToch. *ct, 18 areflex ofIE '0. lE *0 is in fact the most common source of *a: representative cases of *a < *0 can be found in isolated lexical items such as A stwar, B ,4twer 'four' < *kL'etuores and A pctts, B obI. petso 'husband' <,*poti-, in thematic verb forms like 1 pI. A akamas, B ctkem(o) 'w~ lead' < *agorn-. and (cf. aboye) 3 p.I. mido A klyosctntaJ', B klyausentar < *kle7lsontOl', and m, o-grade thematlC (T6I-LO~) nouns like A pl'ClJik, B pl'enke 'island' < *bhronko- (cf. § 33) a~ld A wctr B wel'e 'odor' < *uom-. Since *a from this source regulaTly farls to pal~talize a preceding c~nsona,nt, nothing stan~s in the way?f deriving a form like A rnaskcttal', B rnasketdl' from an earher *rn1}sfcoto1' 0).

Independent confirmation of this is provieled by the speci~l phonological elevelopments associatéel with pl'esents of class IV. Ir-~ th~ bnef excur~us on the Tocharian vowel system which follows (§§ 22---6) rt wlll be our ultllnate object to show that both the characteris~ic root umla~t of these.forms anel theo-timbre ofthe classIV theme vowel m Toch. B pomt uneqmvocaHy to the original presence here, as in class In, of a preelesinential yowel *0.

§ 22. While our knowleelge of Tocharian phonetics is obviously frag.men­tary, it seems fairly clear that vowel length was no~ phonolog!cally elistinctive in either Tocharian language. Thus, there lS no conslstent opposition between i anel i 01' U anel ii in the native lexicon: the infrequent Toch. B graelation seen, e. g., in ptle 'wounel' vs. pI. p-ilénta (Krause--:­Thomas, § 10, Anm. 1) shoulel eithel' be intel'preteel as a pul'ely gl'aphlC imitation of the alternation founel in pairs like áke 'enel', pI. akéntct, 01' as a second, anellikewise subphonemic, alternation ofthe same type. Similal'ly, the miel vowels e anel o are not further subclassifieel accoreling to quantity: Krause anel Thomas' view (50ff.) that the CToch. eliphthongs *ai anel *ct1¿ yieleleel e anel (5 in Toch. A (co~tl'asting in length with. e. anel o Jl'om othel' soul'ces) seems to be purely conJecturaI. Only the OpposltlOn of a : ct , clearly phonemic in both languages, remains a potential case where length is

9) Save where the first vowel is in the root syllable; here -ajii- yields B -oy-, as in soy- (TI) 'be sated' < *sajiija- < *sh2-jejo- (cf. Gk. &ETaL).

10) On the phonology of the first syIlable see § 31.

29

Page 15: Jasanoff.1978.Stative and Middle in Indo-European

phonologieally relevant; here, however, several faets suggest that the underlying eontrast was one of quality rather than quantity. From an etymologieal standpoint CToeh. *a (> AB a) eontinues not only late lE *a and *0', but also the short vowels *a and *¡¡; a further souree is lE *0, whieh, aeeording to the rule diseussed by Cowgill, Studies Lane, 176ff., may be lowered to CToeh. *a before *a in a following syllable (ef. B 01' 'wood', pI. a1'wa < *d01'lla; for other examples see §§ 33, 37 ff. below). Similarly, and with equal indifference to quantity, the CToch. "short" vowel *a continues the lE mid vowels *0 and *e. Given that length is not distinctive elsewhere in Tocharian, and that the normal antecedents of *a and *ií are lE low and mid vowels, respectively, it is attractive to consider the possibility that the Common Tocharian contrast between *a and *a was not one of length but at least partly one ofheight. Under this assumption it would be simplest to regard the historically cognateopposition between a and a in Toch. A, like that between a and e in Toeh. B, as a contrast ofheight also; and since it is likely that (a) had approximately the same phonetic value in Toch. B as in Toch. A, it may be surmised that both A a and B a represent a central mid vowel 11 ).

§ 23. In my view the deseriptive and historie al facts of Tocharian phonology can best be aecounted for by assuming a system like the following for the major vowel phones 12) of Toch. A and B:

i [i] e [e]

a [i] a [A] ¡¡; [a]

u [u] o [o]

Apart from taking "short" a as a mid vowel, this interpretation has only the novelty of treating the "Fremdvokal" éi as the central vowel corresponding to *i and *u. Toch. éi is usually assumed to represent a reduced schwa-like sound. Its exact phonetic properties cannot, of course, be determined, but in several respects it patterns as a high vowel rather than as a mid voweI. Compare such facts as the following: 1) in both languages, and especially in Toch. A, éi is rounded to u, rather than o, in the neighborhood of a velar 01' labiovelar consonant, e. g., AB kulyp- 'desire' < *kwéilyp- (cf. B ger. 1 kwéilypelle beside kulypelle) , Ayuk 'horse' < *yéikwa (B yakwe); 2) similarly, éi is frequently replaced by the high vowel i in the neighborhood of palatal and labial consonants, e. g., A ciñcéi1', céiñcéiJ', B ciñcaJ'e, céiñw1'e 'lovely' < *céiñcéiTa, A 3 pI. -iñc < *-éiñci < *-énti or *-nti, B pilko 'glance' < *péilka (Apéilk); 3) CToch. *éiis not only thereflex odE *e,

11) Tliis ¡s, of course, also the value of a in the Indic writing system from which the Tocharian system derives.

12) On the taxonomic phonemic level, only two heights are distinctive in Toch. B (cf. below).

30

but also frequently of *~I" e. g., A 1'tM, B mt1'e 'red' < *Téit1'a < *1'udhm-, A ckawr, B tkace1' 'daughter' < *téikacaT < *dhng¡¡te1'; 4) in a few instanc~s lE *i appears to yield Toch. *éi after initial *w-, e. g., A wéis, B wase 'pOIson' < *y,iso-, A wéit, B wate 'second' < *dy,ito-.

Under this system the Toch. B alternation between the stressed vowels a and a and their respective unstressed eounterparts a and éi can be interpreted in a phonologically natural way. Toch. B ~as tw~ phone~ic central vowels, one relatively low (a ~ a), the other relatlvely hlgh (Cl ~ a); each has a low and a high allophone, of which the former appears under stress (dke, yákwe) and the latter appears elsewhere (akénta, yéilcwénta). The alternation of a and a thus reflects exactly the same articulatory process as the alternation oía and éi. A eomparably simple statement does not appear to be possible under any other non-quantitative interpretation of the three a-vowels: in this respect, at least, the readings [A] for a and [i] for éi seem preferable to such conceivable alternatives as, e.g., [al] and [8]13).

§ 24. The aboye analysis makes it possible to survey the history of the .lE vowels in Tocharian in a new light. At the outset of the Tochanan development we may as sume the following comparatively straightforward ehanges, of which only 4) is in any way controversia1: 1) palatalization of consonants before front vowels (cf. § 20), 2) passage of *e to *éi and partial development of*~do *éi (§§20, 23)14), 3) merger of*O'and *aas *a, 4) loss of distinctive quality, and 5) lowering of *0 to *a [a] befo re *a in a following syllable (§ 21) 15). Taking *ii as [i], we thus obtain the followingintermediate system:

*i « *'l') *e « *8)

*a«*e;*u)

*¡¡; « *it (*8), *0; *0)

*u « *ü) *0 « *0)

The least well-understood developments in the history of Tocharian vocalism center about the subsequent treatment of the three non-high vowels *a, *e, *0. The normal outeome of *a, as we have seen, is AB a, but neither language shows a in such forms as A pmwr, B pmcer < *bh1'ate1'; A Wallt, B ost 'house' < *llast~l; A poke, B obI. polcai 'arm' < *bhagh-. Similarly, pre-Toch. *e « *e) and *0 « *0) normally fall together as ¿Toch. *a (>A a, Be), but both languages showo, apparentlypointing to CToch *0, in cases like A olcéit, B okt 'eight' < *olctO', AB 01' 'wood' < *dor'~l,

13) This does not, of course, require us to assume [A] as the value of CToch. *a,; cf. below.

14) Schindler has called my attention to the fact that the best examples of the sound change u -+a occur in the neighborhood of a dental consonant.

15) There are, of course, other possible orderings. 2) could as easily have preceded as followed 3); 5), appropriately reformulated, could have preceded 4).

31

Page 16: Jasanoff.1978.Stative and Middle in Indo-European

AB yok- 'drink' ~ *~g!(h-. It !s generally heId tha.t the aberrant r.eflexe~ of pre-Toch. *a, *e, *0 m these mstances reflect the mfluence of a neIghbormg rounded voweI 01' Iabia.I(ized) consonant, but neither the nature of this "labial umIaut" nor its chronological relationship to the Ulliversally assumed unrounding of *0 to *ií has ever been carefully described.

§ 25. The ulterior history of the pre-Toch. non-high vowels, I would suggest, can best be interpreted in terms of two basic processes, which I shall call primary rmtnding and primar'y unround'ing. (These will be distinguished in the discussion below from the purely Tocn. A processes of 8econdary rounding and 8econdary unrounding.) Theeffect of primary rounding was to convert *a and *e to *0 under conditions which remain to be precisely identified; the most important instances involve 1) rounding of *a after a labial 01' labiovelar consonant, e. g., nomo pI. Apons, B poñc 'all' < *pante8; A pmcar, B procer < *bhrater; A wa<$t, B 08t < *ya8tu (for A a beside B osee below), 2)rounding of*abefore *0 in thenext sylIable, e. g., B onolme 'living creature' < *a1wlmo- (: ana81c-'breathe' [cf. Ved. ániti]; fOl' the suffix compare 8yelme 'sweat' < *8'l¡.,lidolmo-), and 3) rounding of *e before a labial 01' labialized consonant, e. g., AB yok- < *eg¡'h-, and probabIy also AB yom- 'obtain' (: Lat. emere) and A yow-, B yop- 'enter'16).

Primary rounding produced abundant new instances of *0; these were offset, however, by the even more numerous cases in which *0 was lost through a complementary, and perhaps simuItaneous process ofunround­ing. Although the exact conditions are again uncertain, pre-Toch. *0 typicaIly survived in Common Tocharian only when it was supported by a neighboring rounded vowel or labiaIizing consonant. The former is the conditioning environment in AB 01' < *doru, and further examples to be discussed shortly; the latter is the environment, e. g., in A k08ne, B k08 'how much?' < *Po-, B p08tañl"íe 'last' < *p08t- (cf. Lat. p08tremu8), AB y80mo 'aItogether' < *en-80ma, and probably A okéit, B okt < *oictO('Il) 17). Else­where *0 was unrounded to the vowel which we have designated *ií.

The phonetic value of this segment is uncertain. Since *ií normalIy yields alllid central vowel (a) in Toch. A, we might consider thepossibilitythat *ií was already centraIized in Common Tocnarian, and that e, its Toch. B reflex, illustrates a specifically Toch. B sonnd change of [11] to [e]. The difficulty with such an interpretation lies in the fact that the lllerger of *ií

16) The suggestion that these roots originally contained a long vowel is due to Schindler (personal communication). It is entirely possible that primary rounding consisted ofnot one, but two 01' three distinct processes, and that the rounding of a to o before o occurred simultaneously with the lowering of o to a before a.

17) Here and in a number of other cases, k appears to have inhibited primary unrounding. Compare Krause~Thomas' statement (p. 64) "Toch. k zeigt anschei­nend eine N eigung zu labiovelarer Aussprache" and their discussion (pp. 49~50) of the Toch. A change of ato 7l in the neighborhood of a velar.

32

with pre-Toch. *e « lE *e) is apparentIy of CToch. date: except in cases where primary unrounding was inhibited, lE *e and *0 have identical reflexes in both languages. If we as sume [A] as the value of CToch. *ií, we must therefore suppose that pre-Toch. *e was first centralized to *ií in the common period, and then fronted again to e in Toch. B. This is not an especially natural sequence of deveIopments, and may rather indicate that the original phonetic value of *ií is directIy preserved, not in Toch. A a, but in Toch. B e. If *ií was in facta front mid vowel, the unrounding of pre­Toch. *0 to *ií wouId simpIy have been a phonetic change of [o] to [e]; in Toch. A this [e], continuing both pre-Toch. *e and unrounded *0, would subsequently have been centralized to a [A]18).

§26. Accordingly, the- vowel system of Common Tocharian lllay be reconstructed as follows 19):

*i [i] *a [i] *u [u] *ií [e] *0 [o]

*ii [a]

In principIe, of course, this system is identical to that of § 24, but the distribution ofthe non-high voweIs is no longer the same: pre-Toch. *ii and *e have become *0 in primary rounding environments, while in primary unrounding environments pre-Toch. *0 has been fronted to *ií.

The CToch. voweIs are preserved essentiallyintact in Toch. B, where the only major subsequent development was the introduction of a [A] as the stressed variant of a and the unstressed variant of a. In Toch. A, however, the inherited system was affected by several important new changes: not only was CToch. *ií [e] regularly centralized to a [A] (a new e, of course, was supplied by the lllonophthongization of *ai), but CToch *ií and *0 underwent a second pair of rounding assimilation rules. Unlike the corresponding processes in Common Tocharian, secondary rounding and unrounding in Toch. A appear in the great majority of cases to have been conditioned exclusively by the environment to the right of the affected voweI. Characteristic exampIes of CToch. *ií > A o are A ñom 'name' vs. B

18) Whether we reconstruct [e] 01' [A] as the phonetic value of*a depends in large part on our interpretation of the first vowel in forms like AB epe '01", AB yerpe 'disk', etc. If the correspondence A e = Be indicates a distinct Common Tocharían vowel *e, ítwouldclearly bedesirable toregard *aas [A] rather than [e]; on the other hand, it is also possíble that A e in forms like these represents C'roch. *a which, for obscure reasons, failed to undergo centralization to a [A].

19) 01', if we take *a as [A] and reconstruct a separate *e for cases like AB epe,

*i [i] *ii [4] *u [u] *e [e] *a [A] *0 [o]

*a: [a]

33

Page 17: Jasanoff.1978.Stative and Middle in Indo-European

~-~---------------------......... ......,.,---------------------re

ñem, A pI. lcowi 'cows' vs. B obI. pI. lcewa'f[/" A onlc 'man' vs. B enlcwe, and, with regressive assimilation, A cmol vs. B camel; compare also the development of a to u in A yulc vs. B yalcwe, A tunTe 'love' vs. B tmilcw and similar forms. The corresponding rule of secondary unrounding converted CToch. *0 to Toch. A a; whether *0 first became *ií [e], which was then subjects to the centralization rule, or whether *0 was unrounded directly to a [A] is impossible to determine. Examples of *0 > a include A pracar vs. B procer, A wa~t vs. B ost, A praslci 'fear' vs. B proslciye, and A onlcala·n¿ 'elephant' vs. B onlcolmo. Here too belong the Toch. A class IV presents, to which we may now return.

§ 27 . We are at last in a position to understand the complex phonological behavior of the fourth present class, which illustrates the interplay of several ofthe rules just discussed. A representative member ofthis class is AB as- 'turn dry' , 3 sg. A asatar, B osotar. This ver b is clearly related to Lat. arere and Gk. &~w, as well as to nominal forms such as Skt. asa- 'ashes' and Gmc. *aslcon- 'ash', all ofwhich are referred by Pokorny, IEW, 68, to an lE root *tís- 'dry'. Whether long 01' short, the initial vowel of a late lE *tís­would eventually have yielded pre-Toch. *a, and *as- is in fact directly recoverable in such extra-present forms as A subj. 3 sg. asa~, B causo 3 sg. asa~~a'f[/" etc. In the presents A asatar and B osotar, however, the root vowel evidently continues CToch. *0, which has been secondarily unrounded to a. in Toch. A. CToch. *osota1', the immediate antecedent of the attested forms, in turn owes its initial vocalism to the primary rounding of *a to *0; the environment for this change must have been provided by the *0 of the following syllable (cí. B onolme< *anolmo-, § 25), which, having no possible secondary explanation in terms of the rounding processes described aboye, can only continue lE *0. The pre-Toch. source of CToch. *osotdr must therefore be reconstructed as *asotar.

It is notable, though hardly surprising, that the sequence *0 . .. *0 seems not to have been subject to primary unrounding in Common Tocharian 20). The phonetic explana tion for this is doubtless to be sought in the reinforcing effect of each rounded yowel upon the other; a partly comparable phenomenon, as Watkins has pointed out to me, is the retention ofmedial -a- in Latin under the influence of an a in the preceding syllable, as, e. g., in gen. sg. anatis, carcaris, etc.

The class IV presents with root vocalisms *ai and *au require further comment. Since presents like A potatdr, B pautotar 'flatters' and B laitotar 'falls' apparently point to CToch. *patltotar and *laitotar, respectively, no obvious phonetic condition will explain the failure of the theme vowel in

20)The only apparent counterexamples are o-grade thematic nouns of the type A pm'lik, B pre'like < CToch. *pra'lika < *bhronko8. Here, however, unrounding would have been phonologically regular in case-forms such as the voc. sg. in *-e, dato sg. in *-O'i, instr. sg. in *-0' and loe. sg. and nomo pI. in *-oi (> CToch. *-i), from which it could easily have been generalized to the rest of the paradigm.

34

such forms to undergo primary unrounding. Probably the simplest course is to :ss.ume that the C~och. process which merged the three lE i-diphthongs as at and the lE tl-dlphthongsas *atl was not yet complete at the time of pri~aryunroun?ing, so that at the period when this rule applied the forms WhICh were ultllnately to become *patltotar and *laitotar were in fact *pout?tar and loitotar, with rounded diphthongs. In principIe, the root vocahsm of such forms could either reflect inherited *ou and *oi preserved intact from lE times, 01' inherited *au and *ai (phonemically *atl and *ai in pre-Tocharian), with the first element rounded here, as elsewhere, befare an *0 in the following syllable. In laitotdr, as remarked aboye (§ 18), lE *oi seems the likelier source of the diphthong; in pautotar and the majority of other such forros, the absence of reliable etymological evidence makes a clear choice impossible.

§ 28. The. class IV presents preserve their CToch. appearance essentially u~lal~ered 111 Toch: B. In Toch. A, however, two phonological changes have slgmficantly modlfied their appearance; these are the secondary unround­ing of *0 ... *0, *ou ... *0, and *oi ... *0, to a .. . a, *au( > o) ... a and *ai (> e) ... '. a, and the syncop~e_ of the class IV tlwme vowel before enClings conta111111g the vowel Ta (cí. 1 sg. plantmar> *plantamar, ptcp. pla~tm~'t¡?, >. *plantama'f[/" etc.). The former development is readily intelligi­ble 111 Vlew of the fact that secondary unrounding in Toch. A is characteris­tically triggered by the absence of afollowing labial environment; the stem vowel *-0- would thus regularly have become Toch. A a, and in forms like *osotdr this change could easily have led to the loss ofrounding in the root syllable ~s well. The partial syncope of -a- is likewise phonologically regular, S111ce Toch. A normally eliminates -a- in medial open syIlables when the prece~ing syllable contains a "Vollvokal" (i. e., a non-high vowel) and the foIlow111g syHable a vowel other than a (01' final i). This phenomenon is also observable in class Il (cf. lclyosma1', lclyosma'f[/, = B lclyausemar, lclyausemane); it is not found in class In only because presents of this class show a-, i- 01' u- vocalism in the root syllable 21 ). For further examples and discussion cf. Krause-Thomas, 46f.

§ 29. Both the class nI presents and subjunctives and the class IV presents therefore point to a uniform earlier type with the stem vowel *-0-. The most obvious possible source for this element, as Krause and Thomas implicitly suggest (pp. 200-1, 225), is the o-colored variant of the lE thematic voweI. Although the lE thematic conjugation is almost univer­saHy reconstructed with an obligatory alternation of predesinential *-e­and *-0- (the treatment by Watkins, ldg. Gr. IIl.1, 102f., 213í., is a significant exception), the appearance of thematic paradigms with "persis­tent" *-0- is not unknown in the attested lE languages: in particular, an

21) Note again the tendenc,y of ti to pattern with i and 1l.

35

Page 18: Jasanoff.1978.Stative and Middle in Indo-European

apparent parallel to the class III (IV) conjugational pattern can be found in Germanic, where the forms of the Gothic passive, 1,3sg. bairacla, 2 sg. bairaza, .1 ~ 3 pI. bafrancla '1 a~ carried, etc.' point respectively to earlier *bherotm, bhet'Oso~, *bheronto~. Whether the non-alternation of the thema­tic vowel in these forms constitutes evidence for a distinct lE thematic type 01' merely reflects the secondary generalization of *-0- at the expense of *-e-, as usually assumed, is a problem which will be deferred to §§ 41-u, where . it wiII be discussed in detaiI. In the followíng sections we shall excluslvely be concerned with the distributíon and spread of this inflec­tional pattern in Tocharían.

? 30 .. The Tocharian forms which definitively establish the etymological ldentlty of the stem vowel *-a- (*-0-) with the lE thematic vowel *-0- are the classIII subjunctives. In our survey ofthese forms (§ 16) we have seen that such subjunctives are attested from eight roots, excluding the doubtful case of tsar- 'separate'. Of these, two, tiím- 'be born' (subj. A cmatar, B cmetar) and wal- 'die' (subj. A wlatar) do not with certainty correspond to primary verbal formations in any other lE Ian¡!lIage: tam­has no clear etymology at aH, despite W. Winter's attempt, IF67, 27-8 (1962), to trace it to a root *clhem-, parallel to *clhehF 'put' ; wal- is probably cognat~ to the root of the petrified Luvian participle 1l( wa )lantis' dead' (cf. Cop, üng. 2, 40--2 [1957]) but the further connection ofthese forms with Hitt. wallszi 'strikes' and nominal forms like Lat. 1l0ln-us, Gk. OUA~, etc. is unclear. The remaining six roots aH have clear verbal cognates elsewhere in the family, and it is a remarkable fact that five can be matched with thematic presents in Indo-Iranian:

kiin- 'happen' (Sllbj. A knatiír): cf. Ved. janami '1 beget' and OLat. geno, -ere 'beget, produce', with passive genor, -1 'be pl'oclucecl'.

kas- 'go out' (A *ksatür, B ksetiir): cf. the isolatecl thematic middle participlejásamana- 'exhausted' RV1.112. 6, 7. 68. 8. , nam- 'bend, bow' (B nmetar): cf. Ved. námate 'bows', Av. 3 pI. n'Jmante 'bow' .~. 57. 18. ~he middle forms are usual in Vedic; in Avestan, 3 sg. acto n'Jmmt~ occurs wlth the same meaning.

pak- 'growripe' (A *pkatar, B pketar): cf. Ved.pacati 'cooks', Av. paéaiti 'id.', with further thematic cognates in Lat. coquo, OCS pekq, Lith. kepü (for *pekü), Alb. pjelc '1 bake'. For the intransitive value ofthe middle forms in Tocharian compare pácyate 'cooks (intr.)' RV 1. 135. 8.

tsiik- 'burn (intr. )'(A *tskatiir, B *tsketar): cf. Ved. cláhati 'burn (tr.)', Av. clazaiti 'id.' (neither, however, with intransitive middle forms), further Lith. clegü '1 burn (intr.)', OCS zegq (for *clegq?) 'id.', Alb. cljek '1 burn (tr.)'22).

22) From a formal point ofview, it would also be attractive to add the class III subjunctive of AB nak-, which recalls Ved. 3 pI. nasanti «nas- 'perish'); the latter form, however, is not an indicative, but a subjunctive.

36

In each instance the stem of the class III subjunctive is exactly equatable with the stem of the corresponding thematic present; the syncopated root vowel of the Tocharian forms points to CToch. *ii, which here doubtless continues lE *e. Root-initial palatalization, which would be phonologicaHy regular before an etymological front vowel, is in fact preserved in A cmatar, B cmetar < tiim-, and in the isolated infinitive B ñmets'i < nam-, but elsewhere has been analogically eliminated. Given that the Tocharian subjunctive is known to have incorporated a variety of originalIy indicative forms to which a modal function has secondarily been attached (see especially G. S. Lane, Lg.35, 159-79 [1959]), it seems entirely !latural to conclude that the five class III subjunctives listed aboye simply continue lE thematic presents *génhFejo-, *gl'és-ejo-, *ném-ejo-, *pékY-ejo-, *dhég'th-ejo-. Theeliminationofthesestemsfrom theindicativein Tocharian and their replacement by the more highly marked presents of classes VIII and X 23) recalls the early su bstitu tion of jdyate 'is born' (cf. 011' .. gainethar) for *jánate in Vedic; note also 2 pI. impv.jasyata RV 1. 191. 7 beside jásamana-.

§ 31. The fact that the class III subjunctives continue inherited thematic presents should lead us to expect a significant number of originalIy thematic stems among the class III and IV presents as well. 8uch forms are in fact fairly common, although unlike the c)ass III subjunctives, they consist almost exclusively of old presents in *-skejo-, the suffix ofwhich has been reinterpreted as pa.rt of the root. Clear cases of sk-presents are A p1'Clskatar 'fears' «piirsk-), A méiskatar' (B méisketéir) 'is, is located', B musketiir 'gets lost' and A mloskaiar 'escapes'. All but the last can be etymologically identified: méisk-, as noted in passing in § 21, probably continues an earlier *m?J-slCéjó- (: Gk. ¡ÚVW, Lat. maneí'e, etc.), which would have yielded *méi'í?~sk- in Tocharian, whence *mask- by dissimilation; m71slc- < *mu(s)-slCéjó- appears to contain the root found in Lat. mouere (caus.) and (with an added s-element) 8kt. m1l~náti 'steals'; parsk­is doubtless related to Go. fa1lrhts 'fearful' (personal communicar tion of H. Craig Melchert), although the apparent o-grade of praskatar < *proskotiir (like that of mloskatéir < *mlouskotar) is problemar tic. The deponent presents ofthese verbs are genuine archaisms, continuing the same lE type as Lat. pacíscor, nancíscor, oblí1líscor, etc. and OHitt. uiskitta 'comes', d71s1cis1citta 'rejoices', 2. pI. pais1catt71ma 'you go', etc. Morphologically, their classIII (IV) inflection, with persistent o-color of the thematic vowel, sets them off from the productive and far more numerous sk-presents, predominantly causative, of Toch. B (classIX), which show the normal alternation of *-e- and *-0- before the personal

23) P. Hollifield, in a personal communication, has suggested that intransitive class IX presents such as A pakn¿ú!tar and B sinastiir 'is oppl'essed' represent the normal morphological replacements of earliel' presents in *-je/o-.

37

Page 19: Jasanoff.1978.Stative and Middle in Indo-European

endings (e. g., 3pl. k¿ilpi.iskentiir 'obtain' < *-skontor, but 3 sg. kiilpi.istar < *-sketol·).

§ 32. Fiveverbs whichform classnI (IV) presents, viz., B klalltk- 'turn', B prlltk- be fulfilled', A yutk- 'worry', A wütk- separate, decide' AB siitk­'spread ou t' , end in a sequence -tk, which recurs ou tside classes In and IV in about two dozen additional roots. Since clusters of dental + velar stop are not acceptable terminations for an lE morpheme, past studies of these forms (s~e Lane, JAOS 85,66-73 [1965], for a summary of earlier views) have umformly sought to analyze the group -tk into individual compo­nents; a satisfacto.ry explanation of the second element, however, has yet to be found. 1 am mdebted to Melchert for having drawn my attention to the close morphological parallelism between the roots in -tk and those in -sk: both groups, for example, are disproportionately well-represented in classes nI and IV, and both have the idiosyncracy offorming class VI (-ni.i-) presents with a "split" suffix in 'roch. A (cf. 3 sg. ki.itiinki.ifj 'stands up' < ki.itk-, wastiJikatii1' 'moves' < waslc-). It may therefore be suggested that t~ is .t?e ~egular refl~x in 'rocharian of an lE dental stop + sk; the sllnphflCatlOn of *tsk to tic would then recall the treatment of stop + s + stop clusters in Indic, where s is similarly lost without a trace (cf. aor. 2s~. patthalp [AV] beside 1 sg. patsí, 3sg. ábhakta beside 1 sg. abhakfji) '4).

If correct, this explanation of root-final-tk would suggest a straightfor­ward etymology for the verbs ylltk- and wéitk- and provide a rationale for their class nI, inflection.The present A ylltkatar can plausibly be derived from *jlldh-skóto1', where *jlldh- represents the root of Lith. j1ldéli 'be agitated', Ved. yúdhyati, -te 'fights' (= Av. yüir5iia-) and, with causative mealling, Lat. htbeo. Although no other branch of, the family attests a comparable sk-present, the relationship of *;iudh-skéjó- in Tocharian to *illdh-jéjó- in Indo-Iranian recalls exactly that between *gl'm-sfcéjó- in Indo-Iranian (Ved. gácchati 'goes') and *gY1l~-;iéjó- in Italic (Lat. 'ueniO) and Greek (~dvw); note also the root aorists Ved. ptcp. yodhaná- RV 1.121.8, and Ved. 3 sg. (á)gan, Arm. ekn 'came', Toch. B sem 'id.'25). Similarly, wéitlc-, the.basic meaning ofwhich appears to be 'separate, divide', can readily be denved from a late lE *tlidh-skéjó-. The root is doubtless that of Lat. dTilido, which is in turn usually connected with Ved. vídhyati 'pierces' 26) ; for

24) To be sure, there are no cIear instances -of underlying *-tsk- in Vedic: The morphologicaJly productive outcome of *-t- + *-sk- in Tocharian is *-t-iisk-, wi th epenthetic -ii-, cf. B 3 sg. luta<$<$iirIJ < l'llt- 'drive off, causo 3 sg. s:pantiissiim < s:piint-'trust'. . ..,

25) With e-vocalism, perhaps suggesting the former existence of a lengthened­grade root aorist *rJ!'em-t.

26) IfVed. vid/¡- i~ in turn a zero-grade of *l,li-dhehr (see K. Hoffmann, Spr.15, 1-7 [1969]), the ¡¡ ilsence of a laryngealreflex in wiitk- (for expected *wiitask- < *uí­dh;¡-sk-) would recal! the same phenomenon in Hitt. zikk- [tsk-] < *dh(<J)-sk-. '

38

the phonological development of lE *~i- to CToch. *wii- see § 23 and the examples there cited.

A number of other tk-roots, although no longer associated with thematic presents, invite a corresponding analysis. AB litk- 'fall away, depart' is clearly an extension of AB lit- 'fall down' and B lait- 'id.'; the final cluster points to a present *lit-skéjó-, with the root (*leit-) of 01 líoa 'depart, go' (= Go. ga-leipan 'go't and Av. irifJiieiti 'dies'. Similarly, AB putk- 'divide, distribute' < *put-slcéjó- is related to Lat. p1do, and OCS pytajq '1 investigate', while AB martlc- 'cut,. clip' < *mrd-slcéjó- is probably the Tocharian cognate of Lat. mordeo. To these may al so be added AB néitlc-, which, though glossed 'stützen' by Krause and Thomas, Toch. Elem. II, 110,203, appears often to mean 'hold distant' 01' 'push aside' in Toch. B., cf. Toch. Sprachr. 1, 33, b 2-3 ompalslcoññe piist pranlciifjfjiir(l" natlcar¡~ laulce aisamñe, yarlce peti ñafjtiir sü 'he gives up meditation, holds knowledge distant (Sieg and Siegling: 'halt fern') and craves only honor and flattery'; further 8, b1 (palslcal)ñ(e)nta palslcomer(l, n(iit)lcnallona '(die Vorstellungen) [sind] vom Geiste her zu halten'. It i~ attractive to consider the possibility that this verb represents lE *nud-slcéjó-, with the root of the Indic sixth class present nlldáti, -te 'push, thrust away', root aor. 2 sg. nutthálp RV 6.44. 5. For the phonological development *u > ii cf. § 23.

These etymologies, and others like them to be discussed by Melchert in a forthcoming study, make it probable that the five class nI (IV) verbs in -tic, like the four verbs in -sic discussed in the preceding section, ultimately continue lE thematic presents in *-skejo-.

§ 33. Of the remaining verbs which form class III (IV) presents it is notable that not a single example can plausibly be traced to an lE thematic present of an apophonically normal type. Apart from the sk-presents just discussed, the majority of class nI (IV) verbs whose etymology is known correspond to roots with punctual (aoristic) meaning in Indo-European; in such cases the Tocharian present stem corresponds to a thematic 01' root aorist elsewhere. This pattern is illustrated by the following verbs, listed below in alphabetical order.

AB ar- 'stop' (pres. A amtar, B orotiir): cf. probably Glr. &p'l'O, presupposing unaugmented *iíp'l'O < *(hj)órto, whence also Ved. árta, árta 'went' and Hitt. artat 'stood'. For the semantics compare Gel'. aufhoren and especially Hitt. Icarp-, which in the middle means both 'finish' and 'rise'. The a-vocalism oftheextra-presentforms (e. g., A subj. 3 sg. amfj, Bpret. 3, sg. am) reflects the regular passage of *or'-a- to *ivr-a- in pre-Tocharian; cf. §22.

AB 1c1Il- 'subside, slacken' : the root is perhaps to be equated with that of Lith. gulti 'lie down' (stative gllleti 'lie'), gvat§C:¿as 'stretched out'; note A Ic~/i, B Iclyiye 'woman, wife', with a semantic development similar to that seen in Glr. &AOXOC; (J. Schindler, personal communication). A possible fl1rther comparandum is Glr. aor. mid. ~A~'l'O, the specifically passive, rather

39

Page 20: Jasanoff.1978.Stative and Middle in Indo-European

--------------------------.................. ----------------than ~n~ransitive mea~~g of .which ('was felled') could have arisen by OpposltlOn to the transltlve actIve ~&AAÚ) 'throw' (cf. Lat. iacere: iacere). AH ~hree s~ts of f~rms ca~ b~ referred to a root *gyel(h1)-: the circumflex mtonatlOn of Llt~. gu'{t~ pomts to an ani( root shape in Baltic, while ~A'~TO presum.ably con~mues a se( form *gy¡h1tó. The position of Toch. kul- itself lS amblguous, smce CToch. *lcula- could equally well reflect *gul-ó- 01' *gttlhró-. .

.A p.ark- 'go up' : cf. Arm. aor. ebarj ( < *ebhrghet) < bar'nam 'raise, rise' ,RItt .. lmpv.,parkta:ru <park- 'rise'. The latter form appears to represent an mherlted root aorlst, to which may originally have corresponded a present *bhrgh-jéjó- (> Ritt. par'kiya- 'id. ') .. ,. _ B pr~nk- 'restrain oneself' : the evidentlyrelated noun Aprank, B prenke lsland shows that the sense of the verb is probably a specialization of 'separ~te oneself, cut oneself off from'. The root may be compared with Skt. bhmr(l,s-, bhms-, usually 'fall', but in its earliest non-causative occurrence (as a thematic aorist injunctive) 'fall away from, be removed from': RV 10. 173. 1 má tvát rii<$(rám ádhi bhrasat 'die Rerrschaft soll dir nicht entfallen' (Geldner). We ~ay reconstruct}E *bhrenk-, the abnormal phonological structu:-e of~hlCh s~ows the *-k to be an enlargement. In Indic only the thematlC aOl'lst bhmsat appears to be old; the present bhramsati attested from the Brahmanas, is probably a secondary formation o~ the' model of pairs like (á)vidat: vindáti, (á)mucat: muñcáti, etc. 27 ).

B mars- 'forget': cf. Ved. aor. 2 sg. mid. mr<$(hálp RV 3.33.8,3 pI. mrsánta RV7. 18. 21 'not heed'. .

B lttk:.'shine' (pr~s.lyuketar): cf',Ved. aor. ptcp. ruciiná- 'shining' (+ 1 sg. opto ructya TA), Rltt. pret. luktat .became light'. Note that the Tocharian form ca,rmot be directly equ~~ed with the thematic present represented by Ved. rocate and ~v. ~aoéa~ft: a preform *leukator would have yielded Toch. B * lyauketa?' , wIth the same treatment of the diphthong as in klyau<$.~ar. 'hears' (: Ved. ptcp. sró<$amiina-). The initial palatalization of lyuketar lS probably on the model of forms like the causative preterite lyauksa :< ~Z-euk~s-" where palat~lization was phonologically regular.

AB wtk- vamsh : whether thlS verb should be compared with Ved. visáti 'enters' (= Av. visaite) 01', as Melchert suggests, Ved. vijáte 'rushes away, flees' (: OE wican, ORG wihhan 'weichen') cannot be resolved with cert~inty, but the latter cho~c~ is clearly preferable from a semantic point of Vlew. The two roots exhlblt much the same behavior in Vedic: both preser,-ve direct traces of a middle root aorist (cf. 3 pI. ávisran RV 8.27. 12, 3 sg. vtktaRV 1. 162. 15), and bothhave developed class VI (tudáti) presents from earlier thematic injunctives (cf. below). , A ~ik- 'overflow': cf. Ved .. the~. aor. 3pI. asican, ínj. 2sg. sicalp, etc. pour , Av; class VI preso fra8~éantt Yt.14. 54. A middle root aorist ofVed.

27) The creation of bht'C!1'I!sati may also have been facilitated by the presence of the noun bhra1'l!sa- = Toch. B. pre'like.

40

"-1

sic- appears to bepreservedin sécate RV 10. 96.1, if, as the sense allows, this form is a "short-vowel" subjunctive rather than the thematic indicative assumed by Grassmann, Wb., 1516. The complete absence of a thematic stem séca- elsewhere would clearly favor such an interpretation.

§ 34. Another verb for which the former existen ce of a root 01' thematic aorist may reasonably be ihferred, even though no old aorist is directly quotable from Greek 01' Indo-Iranian, is B lip- 'be( come) left, ü brig bleiben'. The immediate affinities of this word are with 01 lifa 'be left, live' and its cognates Go. liban, ORG leben, OS libbian, O E libban 'live', all pointing to a CGmc. 3 sg. *libaip (class III weak). Both the Tocharian and Germanic forms seem. to reflect lE *leip- 'smear, cause to adhere (act.) j stick to (mid.)', semantically influenced by the phonologically similar *leik11

- 'leave (act.) j be left (mid.)'; the same root also underlies an e-stative in Slavic (OCS pri-lbpéti 'stick to'). A thematic aorist of *leip- is directly attested in OCS pri-lbpe (: preso pri-lb(p)nqti 'become stuck to'), and is indirectly presupposed by the nasal present which appears outside Slavic in Lith. limpu '1 stick' and Ved. limpati (AV) 'smears'. Cf. also Class. Skt. them. aor. alipat, root aor. mid. alipta.

There is likewise reason to suspect that a middle root aorist was originally formed by lE *llag- 'break', which appears in Tocharian as B wiik- 'break open, split (intr.)' (pres. wokotar; cf. subj. V A wiika<$). The perfect of this root, which is seen most clearly in Gle ~~ye: 'is broken', is probably inherited; Ritt. wiiki 'bites', the transitive value of which may be due to the polarizing influence of a competing middle paradigm, is perhaps traceable to a perfect as well (cf. §4). The existence in Indo-European of a perfect *(ye)yiÍge is itself sufficient to suggest the former presence of an intransitive aorist *yag(t)a (cf. § 3). In Greek, *llag(t)O was apparently replaced by.the aorist passive E(f)&Y-1J in the same way that, e. g., ¡.t[XTO was replaced by ¡.t[y"IJ (cf. Chantraine, Gr. ham.I, 400).

§ 35. The class III (IV) presents ofthe ten verbs just discussed would thus seem to be related to root 01' thematic aorists elsewhere. An obvious expedient would be to compare them with the sixth class (tudáti) presents of Indo-lranian; these forms, as has long been recognized, are based on earlier injunctives outwardly indistinguishable from thematic aorists. A direct equation of the two categories, however, although it would account neatly for pairs like B wiketar : Ved. vijáte 01' B pranketar : Ved. bhrasat would not be free of difficulties. The"tudáti-type is well-attested only in Indo-Iranian and Balto-Slavic; both these branches also exhibit a well­developed thematic aoristo The latter formation belongs to a comparative­ly recent, stratum of the lE verbal system: as shown by Cardona, The Thematic Aarists of Indo-Ettrapean (unpublished Yale University disserta­tion, 1961), only two such aorists, *yidét 'saw, found' < *yeid- and *h11udhét 'went' < *hieudh-, are attested in as many as three independent

41

Page 21: Jasanoff.1978.Stative and Middle in Indo-European

lE traditions. While it is true that the thematic aorist has enjoyed considerable productivity in the lE dialect area consisting of Greek, Armenian, Indo-Iranian and Balto-Slavic, it is not nearly so well represen­ted in the more westerly branches of the family. In Tocharian the only attested example of a thematic aorist is A lac, B lac « CToch. *ltlc <*h1ludhet; cf. Gk. ~AUe¡;, 011'. luid 'went'), which serves as the normal preterite to AB lat-, lant- 'go out'. This form is an isolated survival from Indo-European itself; it cannot in any way be taken to justify the assumption of a whole series of thematic aorists at a stage of pre-Tocharian just prior to the emergence of the classlII (IV) presents.

A further consideration is the following. Sixth class presents in Indo­Iranian are characteristically active ra ther than middle, and only exception­aHy media tantum. This is even more strikingly true of the thematic aorÍst : of the nearly sixty roots which attest this formation in Vedic Sanskrit, only fourteen are found with finite middle forms, and for at least five roots these are demonstrably secondary 28). The predilection ofthe thematic aorist for the active voice has nó synchronic motivation in Vedic, and is almost certainly an archaism, while the routine occurrence of second aorist middles in Greek can easily be secondary. Thus, the exclusively middle inflection ofTocharian class III (IV) presents argues against a derivation of these forms from thematic aorists 01' injunctives, which probably originany lacked a middle altogether. On the relationship between thematic and middle root aorists see further § 47.

§ 36. The conclusion suggests itself, therefore, that while the class nI presents surveyed aboye may in individual cases continue preforms which were already thematic in Indo-European, the type as a whole is best regarded as an inner-Tocharian development ofthe lE middle root aorist injunctive. ]rom a formal point ofview it is easy to see how the spread of thematic inflection might have proceeded : the identity ofthe thematic and athematic middle paradigms in the 3 sg. would have provided an obvious starting point for the complete elimination of the latter type. As we have noted obliquely in § 3, the oldest form of the athematic 3 sg. desinence in Indo-European may be reconstructed as *-0 (01', with a following particle, *-01", *-oi; cf. §45); moreover, it is clear from archaic formslike Hitt. neya, neyari 'is led' (: Skt. nayate) , lJ,aliya, lJ,aliyari 'kneels', etc., 011'. pass. 'berar, 'berr 'is carried' and probably sóbhe 'shines' RV 1.120. 5 (= sóbhate) that the

28) Thus, 3 pI. budhánta and mr~ánta are mechanical replacements of *budhra[n] and *mr~ra[ n] ; the ending -raE n] is systematically excluded from the injunctive (cf. Hoffmann, Der Inj7¿nktiv im Veda, 227). Similarly, vidánta, ifnot simply created to supply a middle counterpart to 3pI. V1:dán, may have supplanted *vidra[n]; avidanta and 3 sg. vidata are first found in the Atharvaveda. 3 sg. arata (beside root aoristarta) and 1 pI. aramahi (AV) are based on 3 pI. aranta, thereplacement ofolder *arira[n]. 1 pI. gamemahi appears to have been created as a medial pendant to acto gaméma.

42

same ending must be reconstructed for the theÍnatic conjugation as well. For Tocharian, as, mutatis mutandis, for Indo-Iranian, the simplest course is probably to assume that the occasional aspectual ambiguity ofthe aorist in its non-indicative uses led to the creation of a class ofpresents based on the stem ofthe aorist injunctive; such a development would o bviously have been favored by the pre-Tocharian loss of the aorist as an independent aspectual category. Once established in the present system, a.the.m~tic 3. sg. forms like *bhrghó(r) , *lukó(r) etc. wouldhave been formally mdlstmgmsh­able from thematic presents like *pék!'o(r), *m1}s/Có(r) and *iudhs/Có(r). The paradigm of the latter forms, characterized by "persistent" o-color of the thematic vowel, was ultimately generalized; as an independent and later development the 3 sg. ending *-0(1') itself carne to be replaced by *-otar, whence class In *_Cúar 29

).

§ 37. The presents just discussed represent transformations of t~e lE aorist· and as such are exclusively non-stative in value. We shall see m the follow'ing chapters the significance of th~ fact that, notwiths~an~ing t~eir meaning, it is precisely these forms WhICh are correlated wlth e-statlves elsewhere.

The list of class In (IV) aorist presents can probably be expanded. Thus, the lE roots which underlie AB as- and B palk- form second conjugation statives in Latin (are re, f~tlgere), and despite the absence of independent evidence for late lE root aorists *lís-ó and *bh!g-ó, it is not impossible that such forms provided the starting point for the creation of CToch. *osotar and *palklítar 30

• Here too may belong o~her class III (IV) pre~ent~ wit~ zero-grade of the root, such as A platar goes, out, ,beco~es e~tmgu~she~ (probablyrelated to Lat.flare and OE blawa'~ blow ), Atnwatr:r, B tr~wetar 'mingles' and B sr1lketar 'dies'. The lack of smtable w?rd ~quatlOns ~or these and similar forms, however, makes a clear determmatlOn of thClr status impossible.

29) In his importan t Harvard University dissertation, On the Sy'stem of Oonju;ga­tion in Proto-Indo-E7tropean (1977), Hollifield takes a somewhat dlfferent posltlOn. Hollifield holds (ch. 12) that certain middle root aorists, which originally were characterized by full-grade, ga ve rise within Indo-European itself to deríved mi~dle presents with zero-grade root vocalism; CToch. *wi~átiir a;r!:d Av. visaite, ~.~is Vlew, continue a single inherited present based on the aOrIst *yétk(t)o. The posslblhty that the stems under discussion owe their zero-grade to a process of internal derivation is not unattractive' that such stems were thematic in both voices, and occurred with a full range of pre~ent indicative forms, however, appears doubtfuI. Under either analysis, of course, the connection of the Tocharian forms with the lE middle root aorist is unmistakeable.

~O) In the case of *bhleg- the existence of an lE sk-present (cf. Ved. bhrjjáti 'roasts', Lith. blizgeli 'glitter' , Toch. AB piilsk- :think', all with an analogical voiced cluster for phonologically regular *-sk- < *-g-sk-, as in Gk. [J.(cryw for *[J.(cr)(~) arg,;es for the former existen ce of a root aoristo Tocharian is the only lE language m WhICh the root *lís- underlies a primary verbo

43

Page 22: Jasanoff.1978.Stative and Middle in Indo-European

,-_~_~~ _______________ p.s __ -----------_-.t § 38. Four of the verbs listed in §§ 15, 17 show presents with lengthened grade. As noted earlier, A samatar 'grows' and salpatar 'crosses over, is redeemed' point to preforms with original *-e-; despite their synchronic agreement with classIV, these forms continue CToch. samatar, *salpatar ( < pre-Toch. * semotar, * selpotar) and thus belong historically to class nI. In Toch. B the same roots form class In presents of the regular type (tsmetCir, tsCilpetar); here, however, lengthened-grade presents are formed by nu- 'roar' (ñewetar) and l71- 'send' (lyewetar), which lack equivalents in Toch. A. It seems likely that these forms originally belonged to an lE "acrostatic" paradigm, characterized by lengthened grade in the active singular (cf. Ved. tá~~i (AB) 'fashions' (= Av. tiist'i) , stáúti 'praises') and full­grade elsewhere (cf. 3pI. ták~ati, 3sg. mido stáve)31). The e-vocalism of samatar, salpatar appears to have spread from the (unattested) forms of the . active, while the zero-grade of tsmetar, tsalpetar probably reflects a substitution of zero- for full-grade {jf the same type as that seen, e. g., in Ved. 3 pI. st71vánti for *stávati < *stéy,'i],ti32 ). The vocalism of ñewetii.r invites identification with that of Ved. 3 pI. anavan 'roared' RV 10. 68. 1, but no such comparisons suggest themselves for lyewetar, samatar and salpatar. The latter two verbs correspond to thematic presents of the ordinary (full­grade) type in Gk. aÉ(1.w (cf. especially at(1.OCC; 'bodily form, build'; so Pedersen, Toch. 21 7

) and Lith. telpu, tilpti 'fmd room, enter', respectively; whether these should in turn be regarded as replacements of earlier *dém­mi and *télp-mi is impossible to ascertain from the available evidence.

As 1 shall argue at length elsewhere, there is also considerable reason to believe that Indo-European possessed acrostatic presents of a second type, characterized by o-grade in the active singular and e-grade elsewhere; a representative example is Go. malan 'grind', Lat. molo, -ere beside 011'. melid and OCS met'jq, for which an athematic paradi~m with apophony was correctly posited in 1916 by Meillet, MSL19, 186 3

). It is possible that several class IV presents were originally inflected in this fashion. We have seen in § 18 that the regular Toch. A correspondent of B laitotar 'falls' is litatar (In); it is attractive to suppose that the o-grade of laitotar is basically that of the corresponding active (compare the *-e- of salpatar,

31) See. J. Narten, Pratidanam, 9-19. 32) Under this analysis the four presents in question would constitute our best

examples of class IH (IV) verbs derived from lE athematic presents. The absence of initial palatalization in B tse'liketar 'arises' (probably < *dhengh-, IEW 250; cf. A tsa'likar 'peak') makes its derivation from an acrostatic present less certain.

33) Although the point is only of incidental relevance here, 1 believe it can be shown that presents of this type were characterized not by the regular active endings (*-mi, *-si, *-ti, etc.), but by an archaic form ofthe middle endings similar to those traditionally reconstructed for the perfecto The decisive evidence for this view is provided by Hittite, where the cognates of verbs like maZan and molo typically belong to the (t,i-conjugation (cf. 3 sg. mallai 'crushes', kanki 'hangs': Go. hahip<*hanhip, etc.). See also §§54, 70, 72, 90.

44

I I I

\

samatar) , while litatCir reflects the replaceme~t of.an e~rlier e-grad.e by ~er~­grade (cf. tsalpetar, tsmetür; e-vocalism is mamtamed m Go. ga-le~pan ~o ). Similarly, A tsarwatCir 'is confident: « *torpo-; ?f:, B opto (s~bJ. Vl tsarwoytar) is clearly related to the famIly ofVed. t'(pnot~ becomes satlSfie~ and Gk. TÉp7tO(1.OCL; the assumption of a present *torp-j*terp- would explam the aRophonic relationship of the Tocharian verb to its Greek counter­part3 ). Other class IV verbs which may in 'pr~?ipl~ be~ong here a~~ ~at- :be capable' (A yatatar, B yototar; cf. Ved. yatat~ stnves, Av. yate~t~ begms moving') and B aiw- 'devote oneself (cf. A ru- 'be~d forward:), a,s w~ll a,s such isolated lexical items as A pot-, B paut- flatter and B kla~ks- shnvel .

To be sure) other explanations for these forms are in J?rinciple available. AB tsarw- is stative in meaning, and thus could theoretlCally be compared with the o-grade perfect attested in Go. pa1j'nee~~' (c~ .. h~wever.§ 7~); yiit-, likewise stative, recalls the Avestan perfect yanata lS m motlOn , ptcp. yaat¿s. The verb lait-jlit-, while unlikely on sem~ntic grounds to ~eflect ~n old perfect, conceivably continues a formation like the Indo-Iraman aonst "passive": this category, as S. Insler ha~ shown (IF 73, 31~-:-4.6 [~968]): historically opposed o-vocalism in the smgular (cf. Ved. (a)Jo~~, (a)vart~, etc.) to zero-grade in the plural (áju~ran, áv'(tran).

§ 39. Several class In (IV) presents show occasional instances of active inflection. The active forms listed by Krause and Thomas (200, 202) are the following: A 3 sg. karya~, 3 pI. karyeñc, B? pI. keriyeny, :: A kary-, ~ ke;y­'laugh'; A3 sg. klawa~ < klav;- 'pr~clai~ ; A ptcp:, tnkant <" tn~- be confused' ; A 3 pI. triweñc < tnw- mmgle ; A ptcp. maskant, B maskenca < mask-' A 3 pI. ritweñc, ptcp. ritwant < ritw- 'connect (intr.)'; A 3 sg. 8'parc~~-any" ptcp. sparcwant < spartw- 'turn'. O~ th~se, A sparcw~- (for *sparcwas-) and sparcwant show root-final palatahzatlOn and thus clearly belong to 'class n; they require no furthe~ discussion here 35). Four of t.h~ six remaining verbs have active forms only l~ ~he 3 pI. and present_partlClple. From a historie al point of view the partlClples m A -ant, B -enea are not properly "active" at all: the original indiff~rence ~o voice of.the ~E .verbal adjectives in *-nt- is w~ll~kl1~wl1 from theIr J?asslve value m HlttIte (~f. appant- 'takel1' ,piyant- glven ,etc.) and occaslOnallyelsewhere, as, e. g., m Lat. euiclens, ttehens 'qui est véhiculé', Ved. pf~ant- 'speckled', etc. It therefore seems likely that the association of forms like ~ m~skant an~ B maslceñca with the otherwise deponent class III paradlgm lS a gel1ume

34) Since Tocharian has no "morphological" zero-grade in. *~~- ofthe type fOl~.n~, e. g., in Greek, !talic and Celtic, there is no reasonable posslblhty that tsarwata1' lS based on a root-allomorph *tarp-,

35) Similarly, 1 omit from the líst of active forms the wholly anomalous,?resent.?f A yam- 'make, do' (sg. 1 ypam, 2 yat, 3 yali, etc.) and A 1 pI. tsaTa,rtWS « ts~r­'separate'), which differs in vocalism from B tsretal' (lII) and could Just as easIly belong to class IL

45

Page 23: Jasanoff.1978.Stative and Middle in Indo-European

archaism in Tocharian, and that the few "active" class III participles which survived beyond the CToch. period were themselves the model for the creation in Toch. A of new and phonologically similar 3 pI. forms with the active ending -eñe < *-añe(i).

The only roots with active finite forms outside the 3 pI. are A kary-, B ke1'y- and A klaw-. The latter verb is clearly based on lE *kle1l- 'hear' (Ved. sr;rpóti, etc.), and in both root vocalism and meaning recalls Ved. sraváyati 'makes to hear' and Av. sra1l1laiieiti 'proclaims (the Gathas)'. A direct derivation of Toch. A 3 sg. klawaf? from an lE iterative-causative *kloyéje­would be possible ifwe assumed an early Tocharian contraction of *-eje- to *-e-, as, e. g., in Italic: we might then envisage a phonological development * kloyéie- > pre-Toch. klowe- > CToch. * klaw a- > A klawaf? N othing seems to stand in the way of such an explanation, for although lE *-eje- yields CToch. *-ai- rather than *-a- in A tre, B trai 'three', the treatment of*-eje­in polysyllabic words could easily have been different from that in disyllables 36). It seems attractive, moreover, to attempt a similar analysis for kary-/kery-, which likewise shows o-grade root vocalism and from a semantic point of view cbuld easily be an iterative of lE *(¡her- 'take pleasure, rejoice' (cf. xoc(pe:¡V yÉ:AW-n, Xen. Oyr.8. 1. 33). I would therefore tentatively suggest that lE *-ejo- yielded *-iya- in Common Tocharian (perhaps via the intermediate stages *-iijo- and *-ijo-) ,so that a preform *(¡horejont(i) regularly gave CToch. 3 pI. *kiíriyan(c) and B keriyen;,. Since CToch. *-i- is in any case subject to syncope before y in Toch. A (cf. A opy1ic 'zur Erinnerung' vs. B epiyac), the corresponding Toch. A form (karyeñe) presents no additional difficulties. In the 3 sg., where we should have expected A *karaf? «pre-Toch. *kore- < *(¡horéje-) in place of karyaf?, it is possible that root-final y was generalized from forms like the 3 pI., where it was regularly presento An alternative Toch. A treatment of the sequence *-iya.. is apparently responsible for the contraction seen in the participial form kareman;, (for *karyaman;,; cf. pairs like A yme, B ymiye 'going', A yoke, B yokiye 'thirst', etc.)37).

Thus, the active inflection of class III (IV) finite forms can in every case be attributed to the secondary influence of the present participle 01' to the late and accidental merger of lE *-eje- and *-0-.

§ 40. Our major findings thus far may be summarized as füUows. The classIII (IV) inflectional pattern points to a pre-Tocharian thematic type,

36) If this analysis is correct, B klowotar 'proclaims' would have to represent an analogical replacement of 3 sg. *kleyer¡¡, (: A klawafl). A natural point of departure would have been the subjunctive and preterite stem kliiwa- (cf. subj. V abstr. kliipalñe, pret. 13 sg. kliiwa), to which a deponent present klowo- could have been created on the m,?del of pairs like iira- : orotar, wiika- : wokotar, etc. The apparent replacement of *kloy-eiejo- by *kloy,-ii- in the preterite and subjunctive is archaic; cf. 011'. 'cuirethar 'puts' « *kor-ejejo-), but subj. 'corathar « *kor-ii-).

37) Cf. also subj. V A karefl for *karyafl.

46

confmed to the middle and characterized by persistent (non-alternating) *-0- in predesinential position. Most of the forms which exhibit this inflection in the attested forms of Tocharian continue one of three lE categories: 1) inherited fuU-grade thematic presents, which give the class III subjunctives (§ 30); 2) iterative and inchoative presents in *-ské/ó-, whichgive the classIII (IV) presentsofroots in -sIc and-tk (§§ 31-2); and3) middle root aorists, which give the non-stative class III (IV) presents of roots with zero-grade (§§ 33-7). Other apparent sources of class III (IV) presents include the lE acrostatic present types in *-e- and *-0-, and causatives in *-éie/o- (§§ 38-9).

In arrivmg ato these results we have ignored the problem posed by the persistent thematic *-0- of the forms in question and proceeded on t,!le tacit assumption that invariant thematic stems such as *péPo-, *m'i}skó-, etc. were either secondarily produced by levelling within Tocharian 01' directly inherited from Indo-European. Before bringing Tocharian evidence to bear on the analysis ofthe "e-verbs" ofGermanic and Balto-Slavic, however, we will do well to examine the historical position of persistent *-0- more closely.

APPENDIX: THE THEMATIC MIDDLE IN TOCHARIAN

AND INDO-EUROPEAN

§41. As observed in §29, the non-alternating *-0- of the Tocharian class III (IV) presents and subjunctives has an exact formal parallel in the Gothic passive (1,3 sg. bairada, etc.), which represents the only olear reflex ofthe lE thematic middle in Germanic. These two categories, however, are not entirely isolated: outside Tocharian and Germanic, thematic presents with persistent *-0- are found both in Baltic, where the thematic vowel is uniformly *-a- (*-0-; cf. Lith. 1 sg. vedit 'I lead', 3 p. veda, 2 pI. vedate, 2 duo vedata); and in Hittite, where thematicverbs ofthe lJ,i- conjugation (cf. 2 sg. sarratti, 2 pI. sarratteni, mido 3 sg. sarrattari, 2 pI. sarradd1lma < sarra­'break off, transgress') and, less regularly, the mi-conjugation (cf. 3 sg. tiyazi 'steps', mido 3 sg.lazziyattari 'recovers', pret. 2 sg.lc1lrk1lrislcattati 'you stifled') exhibit the same peculiarity. The Baltic and Hittite forms have been interpreted by Watkins (Idg. Gram. III. 1, 213f.) as evidence for the existence of a thematic type with persistent *-0- in Indo-European; the situation in Tocharian would at first sight seem to lend support to this view.

It may be noted at the outset, however, that the comparative evidence for nón-alternating *-0- is considerably stronger in the middle than in the active. In Germanic and Tocharian the persistent *-0- of the thematic middle contrasts with the alternating thematic vowel of active forms like Go. ba,ira, bairis, bairiP and Toch. A arsam, araf?t, araf?, B ersau, erf?t(o) ,

47

Page 24: Jasanoff.1978.Stative and Middle in Indo-European

~ !

erfiií1'(t « A ar-, B er- 'call forth' (class VIII))38). We shall see directly that traces of a similar distribution of thematic types can be detected in Hittite as well; the agreement of three lE traditions in associating persistent *-0-with the middle voice is too striking an idiosyncrasy to be accidental.

§42. The only branches ofthe family in which p~rsistent *-0- is found in the active are Baltic and Anatolian. It is remarkable that the presence of persistent *-0- in the former group has no parallel in Slavic, where *-e- has instead been generalized from its original domain to the plural and dual forms of the first person (cf. vedem'b, vedevé for *vedom'b, *vedove'). The probable explanation forthis discrepancywas seen m 1891 by G. Uljanov, who noted, Osnovy 21, that in the productive and numerically predominant Balto-Slavic presents in *-jejo- the distribution of e- and o-timbres of the thematic vowel was altered by normal sound change. In Slavic lE *-je- and *-jo-, when not followed by a sonant with which *-e- or *-0- could form a diphthong, ultimately merged as *-je-, so that the appearance of *-e- for *-0- in the 1 pI. and 1 duo of jejo- presents is phonologically regular39). The pattern of the jejo-presents was subsequently adopted by the other thematic present types as well; significantly, however, *-e- did not replace *-0- in the thematic aorist (cf. 1 pI. padom'b, 1 duo padové<pad9, pasti 'fall'), where no functionally equivalent formation in *-jejo- existed to supply the model for such a change.

In Baltic several facts suggest that the absence of e-timbre in the thematic conjugation is not original. Thus, the Common Baltic athematic 2 sg. ending *-sei appears to reflect a blending of inherited *-si with a thematic ending *-ei (cf. Stang, Vergl.Gr. 408), while in Latvian a dialectal 2 pI. in -et survives in the marginal function of an imperative (Endzelin, Lett. Gr. § 671). Accordingly, it is probable that the generalization of e­vocalism to the 1 pI. and 1 du., as we find it in Slavic, represents the Common Balto-Slavic state of affairs, and that the immediate trigger for the replacement of *-e- by *-a- throughout the thematic conjugation in Baltic was provided by the Baltic sound change of *-je- > *-ja-, through which persistent *-a- was regularly introduced into the paradigm of jejo­verbs. In effect, forms like Lith. 3 p. veda, 2 pI. vedate and 2 duo vedata, like OCS 1 pI. vedem'b, 1 duo vedevé may be said to owe their predesinential vocalism to the influence of phonologically regular forms like Lith. lieiia, lieiiate, lieiiata (: lieiti 'lick') and OCS liiem'b, liievé (: lizati 'id.'), respectively. ,

38) Tocharian, of cou:se, also has thematic middles which show the familiar *-elo­:pat~ern, namely, the mlddles of class II and classes VIII- XII, TheÍr position vis­a-VIS the deponents of classesIII and IV is discussed below.

39) 01', more accurately, this is true where the sequence *-jelo- followed a vowel. Mter consonants, where *-je- became *-e- in Balto-Slavic, we must assume that *-i­was analogicallyreintroduced into the 2 sg., 3 sg., 2 pl., etc. before the two branch~s diverged.

48

§43, The reasons for the non-alternation of -a- in the Hittite thematic active are complex, and several classes offorms must be distinguished. The thematic iteratives in -Slc- show no sign of persistent *-0- in the active, where e-timbre is regular in the singular (1 sg. -slcimi, -skinun, 2 sg. -slcisi, -slcis, -slces, 3 sg. -skizzi, -skit) and tends to be generalized to the 1 pI. as well (-slc'iwen( i) beside -slcc¿wen( i), -sgawen( i); the hesitation between 2 pI. -skitten(i) and -slcatten(i) may have its source here). In the middle, on the other hand, -slca- for -slci- is extremely common: texts from all periods attest 3 sg. forms like areslcattal'i 'learns by oracle', karpíslcattari 'lifts, fmished', akkislcattari 'dies', paislcatta 'goes', etc,40).

The conclusion suggests itself that sk-verbs in Hittite were originally conjugated with persistent *-0- in the middle but *-ejo- in the active; the existence of middle forms in -slci- could then símply be attributed to the analogical influence of the -slci-j-slca- alternatíon in the active paradigm. This would recall the sítuation we have observed in Tocharían, where unmotivated deponents like miísk-, m1¿sk- display persistent *-0-, but where sk-presents of the productive type (B class IX) show an alternating thematíc vowel)n both active and midd!e forms (cf. B 3 sg. acto kiilpafifi¿i1?~ 'obtains' < *-ske-, míd. kiílpastiír" < *-sketor).

The productíve and extremely common Hittíte verbs in -iya- (líkewíse mi-conjugation) residually attest the same original dístríbution of -e- and -a-; here, however, ít is generally -a- whích, wíthín the Hittíte tradition, has tended to spread at the expense of -e-o The older language regularly has acto 3 sg. -iezzi, -iet but mido -iyatta(ri), -iyattat(i), while the correspondíng míddle endings wíth e-timbre, -ietta(ri) and -iettat(i), are comparatively rareo It may be noted that the inheríted thematic deponent iyatta(ri) 'goes, marches' (= Ved. áyate according to Hollifield ap1ld Watkins, '1'P S, 1971, 81), shows no sign of ever having had an alternating thematic voweI. The gradual replacement of -iezzi, -iet by -iyazi, -iyat belongs mainly to the historical period and may in part be due to a phonological change of -ije- to -ija- (cf. Carruba, Krat. 7, 157--60 [1960]).

The thematic verbs of the lJ,i-conjugation show persistent *-0- in the active as well as the middle. These verbs are of diverse origins: some, such as gangab}}i '1 hang (tr.)', which appears to show traces of moveable accentuation in Olcl Hittite 41), must have been secondarily thematized at a relatively late date; others perhaps continue thematic presents from Indo­European. In the case of gangalj7}i (3 sg. kanki, -ai), it may be significant that an intransitive miclclle from the same root is attested in Skt. (Br.) sa1ikate 'hesitates'. Since the Hittite analogue of the latter form would have

40) That the sign -kat- is not to be read -kit9- in these forms is shown by occasional spellings such as pa-is-ga-ta-m, VBoT 58 V S. 11.

41) Hollifield has called my attention to the significance of the spelling 3pl. ka-an­ka-an-zi vs. 3 sg. ka-a-an-ki in the Old Hittite ritual for the royal couple, Cat. 2 416, edited by Otten and Soucek (Ein althethitisches Ritual für das Konigspaar, StBot 5).

Page 25: Jasanoff.1978.Stative and Middle in Indo-European

been *ganga(tta)ri, with persistent _a_42), it is not impossible that the

process by which the active forms of gang- were thematized consisted simply in the extension of the invariant stem ganga- from the middle to the active. Other cases in which a thematic f¡,i-verb appears to be based formally on an earlier thematic middle are discussed by Watkins, ldg. Gram. 76-7

43). In sum, the Hittite distribution of persistent -a- in the

active does not seem sufficient to justify the assumption of a corresponding type in lndo-European.

§ 44. Three branches ofIndo-European thus attest thematic middles with persistent *-0-; the historie al status of these forms cannot be determined without a brief examination of the position of the thematic conjugation within the parent language itself. Watkins' groundbreaking study ofthis inflectional class (ldg. Gram. III. 1, passim) has established that the paradigm ofthe lE thematic active may be reconstructed in the singular as 1 *bhéroh2 (= "classical" lE *bhérO), 2 *bhéreth2e, 3 *bhére44 ). The recon­struction *-th2e for the 2 sg. desinence is supported by the Hitt. thematic 2sg. in -atti and Toch. 2sg. A arii~t, B er~t(o) < *-eta; for the 3sg. in *-e cf. Gk. tpÉpe~, Lith. veda, liezia ( < *-e; cf. § 42), OCS vede(t'b) and Toch. A arii~, B er~er(l" both reflecting 3 sg. *o1'se with suffixed particles ofunclear origino A thematic 3 sg. in *-ei (i. e., *-e + the hic et nunc particle *i) is presumably also to be seen in the -i of certainthematic f¡,i-verbs in Hittite45).

In Watkins' view the thematic conjugation had its origin in an earlier athematic type *bhér-h2e, *bhér-th2e, *bhé1'-e; he explains the thematic vowel as the result of a reinterpretation of3 sg. *bhér-e as abare stem form *bhére-~, with subsequent generaliza tion of*-e- ( > *-0-) before laryngeals and resonants) to the rest ofthe paradigm. This is not the place to discuss the arguments-weighty in my judgment-which favor this hypothesis. Whether 01' not we accept Watkins' account of the genesis of the thematic vowel, two facts about the thematic conjugation are of immediate relevance here. First, the endings of the thematic present bear an unmistakeable resemblance to those ofthe lE perfect and middle (cf. §3),

42) The pr~ser.vation ~fthe root-final-k of Skt. sankate, however, makes it likely that thematlzatlOn of thlS form, and also that ofthe hypothetical *gangattari, was a post-IE development. Cf. §72.

43) AIso perhaps a factor in the elimination ofthe eja- alternation was the general predilection of the [¡,i-conjugation for stems with a in their final syIlable.

44) In the 1 sg. Watkins writes *bhéroilo; the reconstruction given here, which was sugges~ed by Holli~eld, accounts more easily for the bimoric *-0 of Germanic (cf. Go. ba¡,ra, OHG bu'u) and the acute *-0 of Baltic (cf. Lith. ved!/', r!;lfl. vedúos). Compare the doublets *-h2e (e. g., in Hitt. eS(Jat(i) '1 sat') and *-hz (e. g., in Ved. áduhi) in the 1 sg. middle.

45) Note that this ending was mostlyretained in Old Hittite, probably because it ceased to be analyzed as -e-i at an early date. The "regular" thematic 3 sg. in -ai is clearly an innovation.

50

and make it likely that forms like *bhé1'e, etc. originally patterned as part of a middle, 01' at any rate diathetically neutral paradigm 46). On the other hand, it is also olear that by the time of the break-up of Indo-European such forms had synchronically been reinterpreted as actives; they are without exception< active in the daughter languages, most ofwhich show a further tendency to replace 3sg. *-e by *-eti (Anatolian (mi-conj.), lndo­Iranian, Armenian, Italic, Celtic, Germanic) andjor 2 sg. *-ethze by *-esi (Anatolian (mi-conj.), Indo-Ira~ian, Italic, Germanic, proba:bly ~reek and Celtic). These endings were eVIdently formed by the sufÍlxatlOn of the particle *i to the corresponding .secondar~ endings *-et an~ *-es; in imitation of the pattern *-s: *-s~, *-t: *-t~ of the at,hematlC actIve; similarly,l sg. *-omi, thoughlesswidelydistributed (cf. Hitt. -iyami, A~m. -em, perhaps in part I-Ir. *-ami), probably rests on the secondary endmg *-om. The very fact that the thematic present *bhéroh2, *bhérethze, *bhére was associated within lndo-European itself with an overtly active imperfectjinjunctive *bhérom, *bhéres, *bhé1'et provides clear proof that *-oh2 , *-eth2e, *-e had been restructed as active endings by the end of the common period47

).

§45. The reinterpretatíon of the original thematic type as an active appears to have led at a comparatively late stage of lndo-European to the introduction of anew thematic middle, the structure ofwhich remains to be examined. In the 3 sg. we have already seen (§ 36) that the oldest proper middle form was probably *bhé1'o(1'), opposed to an active *bhé1'e. The creation of this form reflects a general lE tendency to specialize the ending *-0 in an exclusively middle value, while restricting its apophonic alternant *"':e to synchronically active (including perfect active) functions; the eventual propagation of the 3 sg. middle ending as *-to(1'), rather than *-te(1'), can be regarded as an indirect further consequence of the same development (see below). In the remaining singular forms of the thematic middle there is no evidence to support the assumption of an apophonic contrast between, e. g., *-h20 and *-thzo (middle) and *-hz and *-thze (active). Rather, it appears that outside the third person the formal opposition between the thematic active and thematic middle consisted solely in the fact that the latter was normally characterized by the originallyoptional particle *1'48): the earliest pattern would thus have been sg. 1 acto *bhé1'oh2

46) A systematic functional study of the thematic conjugation in the light of Watkins' reconstruction is undertaken hy Hollifield, op. cit., ch.7.

47) It seem reasonable to speculate that once ~re~. 3 s~. *bhére .h~d b~en reinterpreted as an active, the absence of a formally dlstmct Imperfect¡mJunctlve led to the suffixation of the athematic secondary ending *-t to *bhére in its non­primary functions. The new form in *-et would then have engendered the creation of *-am *-es, etc. elsewhere in the paradigm.

48) F~r ease of exposition, it will be assumed here that *-r was used in Indo­European to mark the primary middle endings of all three persons in the singular,

51

Page 26: Jasanoff.1978.Stative and Middle in Indo-European

---~-----------------~-1

vs. mido *bhéroh2r', 2 acto *bhéTeth2e VS. mido *bhéreth2er, 3 acto *bhére vs. mido *bhérOT, or, after laryngeal coloration and loss, *bhéro: *bhéror, *bhéreta: *bhéJ'etar, *bhére: *bhémr 4D

). This system is nowhere exactly retained, although its essential features are preserved in Hitt. -alJ}.Ji : -aljljm'i, -atti : -attaTi, -i : -ario

§46. The majority ofIE languagespresent the paradigm ofthe thematic middle in a significantly altered formo The middle desinences themselves have been subject to far-reaching changes: 1 sg. endings like Gk. -OfL()(~, Toch. A -mar, B -emaT, and Pali -ame (replacing Ved. -e), and 2 sg. endings like Ved. -ase, Gk. -E[()"]()(~ and Go. -aza reflect the influence of the corresponding actives in *-mi and *_si 50

). The most widespread innovation ofthiskind lS the replacement of3 sg. *bhémr by forms such as Ved. bhámte, GIL cptpET()(~ (for -TO~) and Phryg. ()(~~EPETOp (cf. also Hitt. alclcis1citt((jri 'dies' and MW impers. lcymyscetor « *-etor + i?) 'is mixed'), which show the addition of the productive allomorph of the 3 sg. middle ending to *bhére-, the regular shape of the thematic stem before obstruents. In !talic and Celtic there seems also to have aTisen a 3 sg. ending of the form *-etr- V, which appears, e. g., in Osc. ur:ncteT 'is conquered' and 011'. do' moinethaT 'thinks', with medial-e- preserved. A desinence of this type is pro bably the SOUl'ce ofCToch. *-(Ci)tür as well. From a purelyphonological point ofview, Osc.-Umbr. -ter, 011'. (non-syncopating) -tha(i)r' and Toch. -tCiT can most easily be referred to preforms in *-t1"i; we shall not, however, pursue the consequences of this observation here 51

).

The second majar respect in which the paradigm of the thematic middle has been altered in the daughter languages is in the distribution of the e­and o-variants of the thematic voweI. Under the reconstruction just proposed, the only lE middle forms in which the thematic vowel appeared as *-e- were the 2 sg. *bhéretar and 2 pI. *bhéredhJ.w; in the 3 sg, the *-0- of *bhéroT contrasted with the *-e of *bhére, and in the dual su eh forms as

and of the third person in the plural. The true state of affairs may have been more complex: inSpr. 23, 159ff. (1977) Ihave suggested that *-r, 01' anextended variant *-rí, was originally employed only in the third person, while the *i ofthe hie et nune was utilized for the same purpose in the 1 sg. and 2 sg. Fortunately, the problem of the r-endings is of no immediate relevance to the present discussion.

49) In the 1 pI. and 2pI., on the other hand, late Indo-European had probably already introduced a more Ílmdamental contrast between the active and middle endings. The actualreconstruction ofthe 1 pI. and 2 pI. middle desinences, of course, is problematic.

50) Similarly the 3 pI. in *-onto(r) has presumahl~' replaced an older form containing the desinence found in Ved. d1¿hré, sére, etc. In the 3 pI. of the thema tic conjugation an r-ending of this type is found in Middle Indic, where it is doubtless secondary.

51) See the reference citec1 in note 48; it seems at least possible that forms of the type *bheretn: or *bherotri were created from earlier *blterori by mechanicaIly inserting *-t- after the thematic vowel.

52

Skt. 2 duo bharethe, 3 duo bharete and Toch. B 2 duo impv. pyamttsait 'make!' show that the original predesinential vocalism of the 2 duo and 3 duo was *-o(i)-. These circumstances would obviously have been highly favorable to the establishment of a new 2 sg. *bhéToth2er and 2 pI. *bhémdhye, the effect of which would have been to complete the creation of a thematic middle paradigm with persistent *-0-.

Whether the generalization of *-0- throughout the middle in Anatolian, Germanic and Tocharian reflects an innovation within late Common Indo­European 01' an independent development in each ofthe three branches in question cannot easily be determined. In the former case we would have to assume that the regular alternation of *-e- ancl *-0- which characterizes the thematic middle in languages like Greek, !talic and Celtic is the result of a secondary re-assimilation to the pattern of the active. This would not in itself be impossible, since we have seen that in Hittite the alternating middle paradigm of the s1c-verbs is probably not original, and that in Tocharian persistent *-0- has been eliminated from the middle in present classes where both active and middle forms occur (§ 43). Nothing, however, stands in the way of the alternative hypothesis, under which we would suppose that Anatolian, Germanic and Tocharian extended o-timbre to the 2 sg. and 2 pI. in the post-lE period, but at a time when the 3 sg. ending was still *-or'. (The replacement of *bhémr by *bhéreto1' , of course, would effectively have prevented the establishment of a paradigm with persistent *-0-.) In point of fact, *-or (*-oi) seems to have survived for a significant time in all three branches, at least in certain functions; it is still preserved in Hittite, and its former existence in Germanic and Tocharian can be independently inferred from such facts as the merger of thematic and athematic middles in Tocharian (§ 36), and the identity ofthe 1 sg. and 3 sg. passive in *-adai in Germanic, presupposing an earlier identity of 1 sg. *-ai (=Ved. -e)and 3sg. *-ai < *-oi (cf. Watkins, op. cit., 138).

§ 47. The eVldently secondary character of the opposition *bhére (act.) vs. *bhéro(T) (mid.) suggests the possibility that at an eady stage of Indo­European the choice of *-e 01' *-0 as a 3 sg. ending was determined not by functional, but by phonological criteria which were subsequently obscured. Although the following remarks are necessarily speculative, it may be appropriate to call attention to the apparent relationship between the occurrence of *-e 01' *-0 in the 3 sg. on the one hand and the position of the lE accent on the other. Under Watkins' assumptions, an lE 3 sg. in *-e can be reconstructed in at least three important groups of forms: full-grade thematic presents (*bhére, cpÉpE-~, etc.), perfects (*1lóide, 013E, etc.) and thematic ("short-vowel") subjunctives (*(h1 )ése,Ved. ása[ti] , Lat. eTi[t]; cf, Cret. 3E~){()"E-~, ()(3~X"~()"E-~, etc.). All of these are characterized by fuU grade of the root and barytone accentuation. The 3 sg. middle in *-0, and with it the later ending *-to, on the other hand, is subject to no such distributional constraints, but occurs freely in both barytone and oxytone forms (cf. Ved.

53

I

Page 27: Jasanoff.1978.Stative and Middle in Indo-European

saye but duhé, cá<$(e but brilté, *árta but *mrtá). The hypothesis thus suggests itself that *-0 was originally confmed to forms in which it was accented, and that only after its morphological specialization as the 801e mark of the 3 sg. middle was it extended to other environments52). We would then assume, e. g., *dhughó, *kéjo and *bhéro as the roost aneient reconstructable prototypes of Ved. duhé, sáye and bhára[ti]; an early lE sound change of unaecented *-0 to *-e would have led direetly to the replacement of the latter two forms by *kéje and *bhére53). These were subsequently treated in different ways, *kéje reroaining synchronically a middle and undergoing renewal to * kéjo, and *bhére showing a roorphologi­cal split into two forros, an active *bhére and a middle*bhéro.

The only important exception to the distributional rule just diseussed is the 3 sg. of the thematic aorist, a form normally reeonstrueted with the ending *_ét M

). According to the interpretation of the thematie aorist set forth by Watkins, op. cit. § 83, and supported with further data by Peters, Spr. 21,37-42 (1975), a form such as lE *yidét ( = Ved. ávidat, Gk. é(.f)L8r::, Arm. egit) is ultimately to be regarded as a transformation of an earlier middle root aorist, standing in the same relationship to an underlying 3 sg. *ltidé as Ved. 3 sg. áduhat 'milked' and ásctyat 'lay' stand in relation to earlier áduha (MS3, 3,4; 4, 2, 2) and *ásaya (ef. Waekernagel, Kl. Sckr., 498-502). While the formal possibility of sueh an analysis may be taken for granted, the morphological status of the hypothetieal *yidé is unelear. Hollifield has recently shown (op. cit., ch. 11) that in Indo-European, middle root aorists denoting entry into a state were originally eharaeteri­zed by e-grade root vocalism; the replaeement of sueh aoristsby equivalent zero-grade forms must be attributed partly to late Indo-European and partly to the dialectal period (cf. Ved. ámata 'remerobered' beside GAv. manta). lE *ye'id- was evidently a root of this type; the eOlTesponding stative is attested in the perfect *yóide 'knows'. The oldest forro of the 3 sg. aorist middle of *yeid- is thus perhaps to be reeonstrueted as *yéide, with the regular post-tonie desinenee *-e. Itwas possibly to sueh a form, after the synchronic reinterpretation of *-e as an active ending, that *-t was

52) I of course assume that at the stage ofIndo-European when the distribution of *-e and *-0 was phonologically determined, *-r (01' *-ri) was not yet obligatory in the presen t middle endings (this is still the case in Hittite). There can be no question of an lE rule *'-01' -+ *'-e1·.

53) 01', m~ttati8 mutandis, we could consider the theoretical possibility that early lE *-é yielded later *-ó. Likewise speaking in favor of an accentually-linked rule is the ending *-e of the originally enclitic thematic vocative, and the enclitic particle *-kl'e 'and' (vs., e.g., stressed *kuód 'what?').

54) Note also stressed 3 sg. *-ské, *-jé; here, however, it is possible that the attested active forms were _created later than the corresponding middles in 3 sg. *-skó(1'), *-jÓ(1·).

54

.. I

added 55); the apophonic reduction of *yéidet to *yidét would then have been a separate and later development. Other early thematic aorists may have arisen in the same way: in particular, the association of lE *h1ludhét 'went' (= Gl\:. ~AuSe) with an archaic perfect *(h1le)h1lóudhe (= Gk. dA~AOU­Se) makes it probable that this root too originally formed a middle root aorist with full-grade.

55) Note thatIE *llidét is unequivocally active in function as well as morphology. In lndo-lranian and Armenian *llidét furnishes the active aorist to an active nasal presentmeaning 'find' (Av. vinasti, Ved. vindáti, Arm. gtanem); its valuein Greekis substantially the, same ('see' in the sense 'catch sight of', 'lay eyes on', etc.), although thé present *llinédmi is no longer preserved.

55

Page 28: Jasanoff.1978.Stative and Middle in Indo-European

'" ~

~~ 8 '" '" ::[ '" ~ ~

:g~~ ..o ..o ~ ~..o ..o

~ ..o ..o ~

~ ..o ..o o ~ ~ ~ ~ ~ ~ ~ ~ ~ ~

~~~ ~ ~ '--J ~ ~~ ~

III

THE THIRD WEAK CLASS IN GERMANIO "" ~ .",

'" '" ~ I I ~. '" 8 8 ;:> ~" ~" '"

I ~ ~

Among the traditional subdivisions ofthe Germap.ic verb, the third .", .", .", ~" ~

..o .", ."" §48. ..o ..o ..o ~ ..o ..o if2 ..o .c .c ..o ..o .c ..o ~ ..o ..o

weak class presents the greatest variety of forms fromlanguage to language '" ~ ~ '" '" ~ ~ ."" '" '" o .~ ..s:::~ ~ ~ ~ ~ <::>:,~ ~ and the greatest difficulties of comparative reconstruction. The basic data

'"O may be illustrated with the forms in Gothic, Old Icelandic, 0lc1 High O)

I~ N

German. Old ~axon amI Old English of the verh 'have' 1) (see page 57). I~ I~ ~

I Itl .", El I~ I~ """ I~"

I ~ '" """ I~ I~ "§ I~ " §4\:J. The morphological behavior of Go. haban is entirely typical of its 0 I~ Itl Itl ","

'" Itl o P:i ..o "§ ..o ..o "§"§ "§"§ ..o ..o ~"§ ..o ~

class anc1 calls for no separate c1iscussion. In the other early Germanic ~ ~ ~ ~ ~ ~ ro O ~ ~ ~ ~ ~~ ~~ ~ ~ <::>:,~ ~ .S languages, however, 'have' and several other verbs display features which Ul require specialnote. El " In 0lc1 Icelanc1ic the regular class III type is represented by vctka 'be ..;:

awake' , fro:m which ha fa diverges in 1) showing umlaut in the singular of the e-J--... -¡:-- ~ O) Ul

.", .",

" present (hef(i) , hef(i)r') , 2) lacking a final vowel in the impv. 2 sg. (ha!), and N: ~ N: o ~ ~ Z 3) lacking a union vowel in the inflected form of the past participle (hafar ~ ~ ~ +>

Ul for *hafaar). A further peculiarity is the occurrence of short forms in the O)

singular of the present indicative (hef, hefr); these are archaisms confined to f$ H h olc1er poetry and have no analogue in verbs of the vaka-type. Certain of the O " .", ro

ic1iosyncrasies of hafa do; however, recur in segja 'say' arid pegja 'be silent' : ~ " "., ~ ~

O)

~ "., ~ "., ¡:: ~ segja has a short imperative (seg) and past participle (sagar) , as well as .",

~~~ ~ ~~~ ~ 'h ~~ ~ '"O occasional short forms in the inc1icative singular (seg, segr), but differs from '" o- ~ ~ ~ ~ ~ ~ ~ ~ ~~~ ~~ ~ ~ ~~~

O)

.t; both hafa and vaka in showing an umlauted root vowel, together with an " O) underlying stem-final -j- throughout the present (1 pI. segj1trn, 3 pI. segja, ~ preso ptcp. segjandi); pegja follows segja but lacks short forms in the ..<:1

.~] ~ "" "" H singular indicative and has a disyllabic impv. 2 sg. pegi. "" .~ § "" ~ .~] 8 0)" .", .", .", .", .",

In Old High German the normal paradig.¡n of haben, like that of Go. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ " o ..o ..o ..o ..o ..o ..o ..o ..o ..o ..o ..o ..o ..o O) h ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ..<:1..<:1 haban, is indistinguishable from that of other class III ve1'bs. Along with 0 ~ ~ ~ ~ ~~ ~ ~ ~ ~ ~~ ~ ~ ~ sagen 'say' and leben 'live', however, haben displays occasional fo1'ms of a

Ul " a3~ quite diffel'ent type, viz. 1) 1 sg. preso hab1t, sagu (Tatian, scribe y), as iffl'om '"0..<:1 strong *haban, *sagan, 2) 2 sg. hebis(t), segist, 3 sg. hebit, segit, libit, as iffl'om ~+>

ro f5 *haban, *sagan, *lebctn 01' class 1 weak *hebben, *seggen, *libben, 3) pret. ~ " hapta (Isidore, Monsee Fragments), without a union vowel, and 4) pret. .... C'l M -C'l M C'lM C'l - P, P, <+-i ~:.a

~ "a bÍJ bÍJ ~ .E .E .s _3 w. w. p.,p., ":'1 Q)

u:i ~ po: ~ " 1) For the data in this chapter 1 have relied heavily on the classic study by "" u:i H <D p., p., <D w. H. Flasdieck, Untel'such'llngen über die gel'nwnischen schwachen Vel'ben JJJ. [{lasse, 1'< o .S 1'< ce <D p., p., 1'<

7i11ter besondel'el' Ber7lcksichtig¡¿ng des Altenglischen ( = Anglia 59 [1935J). p., p.,

56 57

Page 29: Jasanoff.1978.Stative and Middle in Indo-European

hebita, segita, libita, past ptcp. gihebit, gisegit, gilibit, again as if from class 1. The Old Saxon paradigm given aboye is the usual one in the ve1'sion of the

Heliand p1'ese1'ved in the Monacensis MS (M). The dialectally distinct and slightly late1' Cotton MS (C), howeve1', has 2, 3 sg. habis, -id, sagis, -id ( < seggian 'say') and impv. 2 sg. habi, sagi; the unexpected absence of i­umlaut in these fo1'ms (but note 2 sg. segis (M) (1 x )) 1'ecu1's in pI. habbiad and iní. habbian (M), as well as in 1 sg. habbiu (C) (1 x). The ve1'b libbian 'live' shows a partial transfe1' to the second weak class, cí. 3 sg. libod, leóot, lebot, pI. libbiod (= libod x *libbiad). .

The g1'eatest diversity of fo1'ms is found in Old English. Beside the p1'operly West Saxon 2 sg. hoejst, 3 sg. hoejp, Anglian poetry and Mercian show the classIl forms hajas(t) and hajap (cí. WS impv. haja); correspon­dingly, secJan 'say' has 2 sg. saJas(t) , 3 sg. saJap. ClassIl forms of libban 'live' are found in West Saxon as well (cí. 2sg. liojas, 3sg. liojap, later leojas(t) , leojap). In Anglian and Kentish 'live' has a distinctive inflection in which libb- is replaced by lifJ-, yielding forms like (disyllabic) iní. lifJan, 3 pI. lifJap; the same morphological type is also attested in a num ber of originally class nI verbs which have otherwise been absorbed into class n. The Northumbrian equivalents of habban, sec3an and libban have the aberrant 1 sg. forms hajo, soeJo and lijo, as iffrom strong verbs. For hycJan 'ponder' see §§ 60, 77.

§ 50. Class nI weak verbs are comlllon in Old Righ German, where they constitute a productive denominative type, but elsewhere their number is restricted. Fewer than thirty verbs inflect like haban in Gothic, and fewer than sixty like vaka in Old Icelandic; in Old Saxon and Old English only 'have', 'say' and 'live' consistently show class In forms, although a small number of other verbs preserve more 01' less substantial traces of an earlier paradigm lilm that of lifJan in the Anglian dialects of Old English. In both Old Saxon and Old English the great majority of original class nI verbs have been transferred to the o-class (classIl).

Synchronically primary class In ver bs are prevailingly stative, although a num ber are simply durative in meaning; semantically typical examples are Go. hahan, ORG hangen 'hang (intr.)', Go. *wunan (in unwunands 'unsatisfied'), 01 una, ORG wonen 'be satisfied', Go. wakan, 01 vaka, ORG wahhen 'be awake', 01 bija, ORG biben, OE bijian (with class III traces; cf. Flasdieck, 45) 'trem ble'. A number of such verbs can be compared directly with Latin statives in -e- andjor Balto-Slavic presents in *-~- with infinitives in *-e- (cí. § 7), e. g., Go. pahan, -aip, ORG dagen 'be silent': Lat. tacere, Go. munan, -aiJ) 'intend, be lllinded' (cí. ORG jir-monen 'despise'): Lith. mine%i, OCS mbneti 'think, relllember' (cí. also Gk. fJ.av~vaL), Go. pulan, -aip, 01 pola, -ir, ORG doren 'endure, suffer': Lith. tyle%i 'be silent', Go. witan, -aip 'observe', ORG gi-wizzen 'be capable': Lat. uidere, perhaps also OCS videt'i 'see', Lith. pavyde%i 'envy' (but see § 90), etc. Pairs of this kind constitute the principal evidence for positing a direct formal relationship

58

between the Germanic forms and the Italic and Balto-Slavic stative types in *-e:-.

Similarly, the dominant group of class III denominatives invites compar­ison wi th the lE sta tive jinchoa tive type seen in La t . .senere, Lith. sene1i, -e(ja and OCS sta1'l!ti, -éjet'b (cí. § 6). Characteristic examples are Go. jastan, -aip, ORG jasten 'fast', Go. ana-silan, -aip 'fall silent' (: Lat. silere) , 01 skorta, OE scortian (cf. Flasdieck, 60) 'lack'; the productive Old Righ German inchoatives in -en, e. g., roten 'turnred' (: Lat. rube're, Russ. Ch. SI. ndeti s~, etc.), blinden 'grow blind' (= OE blindian, with secondary transfer to classn) 'grow blind', tagen 'become day' belong here as welI. A much smaller number of class In denominatives are neither stative nor incho­ative but fa<;ltitive in meaning; such forms are especially conspicuous in Gothic, cí. piwan, -aip 'make subject', weihan, -a1) 'sanctify', jastCln, -Clip 'hold fast' (beside 'fast'), .sweran, -aip 'honor', etc.

§51. It can readily be seen that the classnI p~~'adig~s found in t~e individual Germanic languages are formally classlÍlable llltO three lllalll groups. The forms of Go. haban are synchronically built from two stems: one, haba-, is to all appearances thematic and has the sallle distribution as, e. g., bindCl-, bClira- 01' giba- in the system of the strong verb (hence also passivehabada); the other, habai-, is confined totl~epres .. 2 sg., 3 sg. a.nd2:pI. and to the weak preterite habClida, ptcp. haba~ps. TIllS pattern hkewISe underlies the conjugation of 01 vaka: 2, 3sg. vClkir evidently continues an earlier form in *-eR < *-aiz, while 2 pI. vakio, pret. vClk(}a and ptcp. vokat point respectively to CGmc. *wakaip, *wakauRr and *wakaida-. The remaining Norse forms, with the exception of 1 sg. vaki, are thematic like their Gothic counterparts; vaki itself is an obvious innovation, made by deletingthe -1' of2, 3 sg. vakir on the model of otherparadigms (cf. 1 sg.jliJg (for *jljúg), 2, 3 sg. jliJgr <jljúga 'fly', 1 sg. jer (for *fqr), 2, 3 sg. jerr <jara 'go', etc.). The position ofthe 1 sg. in North Germanic, it should be noted, was already seen by R. Moller, AjdA 20,128 (1894), andrenders complet~ly superfluous the reconstruction of a Gmc. 1 sg. *wakem, proposed by Rlrt, IF 1,204 (1892) as the comlllon source of vaki and Go. waka. .

A second class III type is represented in Old Saxon and Old Enghsh, whiclÍ oppose forms reconstructible with *-ai- (OS habas, -es, habad, -ed, OE hoejst, hoejp, Northumb. hoejest(t) , hoejep) toforms built on a stem in *-ja- (OS hebbian, hebbiu, etc., OE habban, hoebbe). The apparent class Il forms of OS libbian, OE libban (see aboye) do not constitute evidence for an original difference of inflection in this word, but ra ther illustra te the partial transfer of 'live' tothe o-class, a process which has gone to completion in the overwhelming majority of class III verbs in both languages. Silllilarly, the OE (Anglian) lifJan-type appears not to be ancielit: although forms such as 1-3pI.lifJap, opt.lifJe(n), preso ptcp.lifJende, etc. have widely been taken to point to a Germanic stem *libeja- (e. g., by Flasdieck, 158fí.), Cowgill, Lg.35, 13f. (1959) has decisively shown that pre-OE *libeja- is a late

59

Page 30: Jasanoff.1978.Stative and Middle in Indo-European

innovation, ereated by adding *-ja- to the stem *libe- ( < CGme. *libai-) of the 2 sg. and 3 sg. on the model of class 1 verbs like WGme. *domijan 'think', 1 sg. *domijn, 3 sg. *domijJ. Presents of elassII show a parallel, but more widely rustributed remodeling: the type sealfian 'anoint' < *salbojan (for CGme. *salbon) is eommon to all OE dialeets andis sporarueallyrepresented in Old Saxon in forms sueh as ladoian 'invite' (= OE laoian), pI. folgoiad 'follow' (= OE fol3iajJ), ete. 3

).

It is usual to see traces of a Germanie infleetion like that of OS hebbian and OE habban in 01 hafa and segja, the morphological peculiarities of which have widely been attributed to the analogical interaction of inherited stem-forms in *-ai- and *-ja- (thus, e. g., . 2, 3 sg. hefir may continue a form like Go. habais, -aijJ, with umlaut extended from 1 sg. hef < *habjo; 3 pI. segja may reflect *sagjanjJ, with ungeminated -g- from 2, 3 sg. segir < *sagai-). Similarly, the former existence of a pattern *habjo, *habais, etc. in Old Righ German has sometimes been infened from 2 sg. hebis(t), segis(t), 3 sg. hebit, segit, libit (cf. §§ 57-9). Apart from scattered forms like these, however, the third class paradigm in Old Righ German is based on a single, synchronically athematic stem in -e- « *-ai-). This inflection constitutes a third major class III type - one which has no exact counterpart in any other Germanic language.

§ 52. Of the numerous attempts which have been made to explain the morphological behavior of the third class, the overwhelming majority, as noted in ch. 1, have begun with the assumption of an explicit etymological relationship between the Germanic forms and the e-statives of other lE traditions-an assumption which W. Streitberg expressed with the words "jede Erklarung del' germ. Verba, die nicht an die auBergermanischen e­Formen anknüpft, scheint mil' prinzipiell verwerflich" (Urgenn. Gr., 307). The utility of this hypothesis, however, has yet to be demonstrated : as the following survey will show, the lE antecedents of the third weak class remain almost as obscure today as they were a century ag0 4

).

The basic equation of the Germanic forms with the Latin statives in -e­( < -e- 01' *-ejejo-; cf. § 6) was first set forth in detail by G. R. Mahlow in 1879 (Die langen Vokale, 19ff.). By the early twentieth century a variety of theories seeking to derive the third class from an lE type in *-e- (athematic) andjor *-ejejo- (thematic) had already been elaborated; see, e. g., Brug­mann, G1'.2

, 2 3, 203--4 for a representative discussion of the periodo Few

scholars would any longer claim to see evidence for an inherited athematic

3) A similar innovation underlies Alemannic optative forms such as 1,3 sg. salMe, haMe, 2 sg. salMes, habees, etc.

4) l have not, of course, attempted to deal exhaustively with the enormous literature on the third class. FOl' a more detailed survey of older views, the reader is referred especially to the relevant sections of Flasdieck, op. cit., and Wagner, e­Verba,

60

paradigm in ORG habem, -es, -a. From a phonological point of view it is doubtful that lE *-e- would have yielded WGmc. *-e- in this position (cf. § 78), and even if such a development could be justified, a derivation of ORG haba from CGmc. *habep would require us to abandon the attractive equation of habet and Go. habaijJ, the diphthong of which points unam­biguously to Gmc. *-ai-. E. folomé's attempt, Festschr. Pokorny 87-9, !o explain the -ai- of habaiP as a reflex ofGmc. *-e- flies in the face ofGo. faheps 'joy' and is not supported by any other phonological developments in Gothie; an earlier effort in the same direction was made by K. F. Johansson, De den:v. verbo 187.

The overwhelming majority ofinvestigators have therefore preferred to interpret the inflection ofORG haben as the result of a secondaryextension of * habe- from the 2 sg., 3 sg. and 2 pI., where it was the regular reflex of *habai-, to the remaining forms; the invariant stem vowel of the c1ass II verbs (salbom, -os, -ot', etc.) wouldhave supplied an obviousmodel for such a development. The "regular" appearance ofthe ORG paradigm would then reflect a remodeling by whieh forms similar to Go. 1 sg. haba, 1 pI. habam, 3 pl. haband 01' OS 1 sg. hebbi7l, 1-3 pl. hebbiad were replaced by analogical habem, habemes, habent in pre-ORG times.

Athematic preforms in *-e- have been invoked for another purpose as well: since Streitberg, Z7lr germ. Spmchgesch., 73ff., the suggestion has repeatedly been made that Go. 3 pI. haband and preso ptcp, haba.nds owe their -a- to an early shortening of *-e- to *-a- before tautosyllabic resonants (Osthoffs Law), and that -a- was analogically extended from these forms to 1 sg. habcl and 1 pI. habam. Despite its espousal by so recent a source as Krahe, Germ. Sprachw.II, 121, however, the objections to this view are prohibitive. It is highly unlikely that haband can regularly continue a form like *kapgnti: since *-ent- appears to have been shortened to -ind- (via *-ent-) in Go. winds 'wind' < *1.lentó-, an athematic 3 pI. in *-gnt'¿ should rather have yielded *-ind in Gothic 5

). Furthermore, even if a phonological passage of *-enti to -and could be admitted, it is difficult to see why this should have led to the generalization of -a- e1sewhere; the simpler course for speakers ofpre-Germanic would sure1y have been to reintroduce *-e- in the 3 pl. by analogy, thus bringing the inflection of the stem *habe- into agreement with that of class II stems like salbO- (cf. Go. 3pI. salbond)6).

5) This, moreover, is precisely what we should have expected on a priori grounds: in no other lE language is Osthoffs Law accompanied by qualitative, as well as quantitative changes.

6) To be sure, forms like *sa.lbonp may not yet have been contracted from *-ajonti when Osthoffs Law applied to *-enti. But l lmow ofno typological parallel in any lE language fOl' the remodeling of the 1 sg. and 1 pI., and of these fonns only, in response to an Osthoff shOl'tening in the 3 pI.

Note that here and throughout this chapter l have written b, el and g in Common Germanic even where the sounds represented by these symbols were fricatives rather than stops. Similarly, and likewise entirely fol' convenience, l have

61

Page 31: Jasanoff.1978.Stative and Middle in Indo-European

~-~~~~ ~~ ~-~~~~~~-~~--l

According to what is probably the most widespread view, the diphthong of the stem *habai~ continues an originally thematic (or thematized; see below) suffix *-eje-, which yielded fIrst *-eji- and then *-ai- in Germanic. No actual counter-examples to this development are known; as Cowgill has pointed out, however (personal communication; see also Lg.39, 265 [1963]), we should rather have expected that *-eje- would give *-e- in Germanic, just as in class II the comparable sequence *-iije- is contracted to *-0- (via *-oji-). Moreover, whatever the legitimacy of assuming a change of *-eje- > *-ai-, no comparable explanation can be offered for forms like Go. haba, habam, haband, ol' OS hebbiu, hebbiad, which clearly cannot be derived from a thematic stem *kapijo-. (The suffix-fOl'm *-ejo-, to judge from the treatment of *-iijo- (> *-0-) in class II, would almost certainly have yielded a Germanic long vowel, presumably *-e-. Since Common Germanic evidently had no phonemic *_ii_ 7

), the possibility that the -a- of Go. haba, habam, haband represents the result of a contraction can safely be discounted; cf. Flasdieck, 116.) Thus, while a derivation of Gmc. *habaip from lE thematic *kapijeti is not strictly impossible, advocates of this reconstruction have invariably been forced to invoke one or another ad hoc explanation for the Gothic and Old Icelandic forms in *-a- and the Old Saxon and Old English forms in *-ja-. Characteristic expedients have included the assumption of a paradigm with both thematic and athematic forms (e. g., Krahe, loco cit.), the hypothesis of a Germanic type in *-eje-/ *-ejo- (Specht, KZ 62, 77), and the supposition that Go. haba, habam, haband and OS hebbiu, hebbiad are transfer forms from the strong or first weak conjugation (e. g., Brugmann, Gr.2 23

, 203).

§53. As seen in § 12, the West Germanic stem-form *habja- has been compared with stative formations in other branches of Indo-European, most notably the Balto-Slavic present type in *-l:- (cf. Lith. 1 sg. miniu, inf. minéli, OCS mbnJQ, mbneti). This comparison has typically required the assumption of an ablauting morpheme *-ei-/*-'t- (*-iJi-); an element of this shape was first posited in 1891 by Chr. Bartholomae, Stud. z. idg. Spmchgesch. and, partly in association with the theory of lE "disyllabic heavy bases", has continued to figure prominently in discussions of the third weak class to the present day. Variants of the *-ei-/*-l:- theory, of which the most recent has been argued by W. P. Schmid in the work cited aboye, are too numerous to be discussed individually; representative surveys are given by Flasdieck, §§ 68-78 and Wagner, e- Verba, 48ff. In

reconstructed *-an, rather than *-anan , for the Cómmon Germanic infinitive endings, *-@, rather than *-e, for the 3 sg. perfect ending, and (as a rule), *-]J, rather than *-]Ji 01' *-oi, for the ending ofthe 3 sg. present active. Ido not, of course, wish thereby to claim that al! short final syl!ables were lost in the common periodo

7) The first vowel of forms like Go. ]Jahta, ORG diihta 'thought', was, of cOUl'se, nasalized in Common Germanic. ~

62

¡ general, the following features are common to most versions of the theory : 1) forms like Gmc. 3 sg. *habaip are derived either from *-ei- 01' its thematized counterpart *-eje-; in the former case the long diphthong *-ei­must be assumed to have developed differently from other *- V R- sequences in Osthoffs Law environments; 2) Northern West Germanic forms ofthe type *habjo are derived from thematic prototypes in *-jo- < *-(iJ)i-o-; whether the replacement of *-l:- (*iJi-) by *-jo- represents a purely Germanic, or partly lE development is a matter of controversy; 3) the absence of *-j- in Gothic and Old Icelandic (vaka, etc.) is explained secondarily, usually by assuming contamination with the alternate stem in *-ai- (so originally Sievers, PBB 8, 94 [1881]); the apparentretention of*-j­in 01 hef(i) , hef(i)r and throughout the paradigm of segja is taken to represent the survival ofthe original pattern in the two most common, and hence most conservative classIII verbs.

Apart from the questionability of positing a development of *-eje- and *-ei- to Gmc. *-ai- and the implausibility of an analogical replacement of *habjo, etc. by *habo (cf. §61), the *-ei-/*-'t- theory suffers from serious diffIculties at the lE level. From a morphological point of view probably the most acceptable reconstruction of the lE paradigm within the framework of this approach is that of Schmid, who postulates sg.l *kapiimi, etc., pl.l *kapiJimé, 3 *kapiJjénti (op. cit., 83). Yet it is almost inconceivable that such a paradigm, characterized as it is by the completely regular alternation of strong and weak athematic forms, would not have left a trace in Indo-lranian 8), Greek or Hittite (see below); Northern West Germanic, on the other hand, must be assumed not only to have retained the inherited alternation of *-ii- and *-iJj(o)-, but to have extended the weak stem in a wholly unparalleled fashion to the 1 sg. as well. Wagner and others, who as sume a thematic 1 sg. for Indo-European itself, obviously avoid the latter difficulty, but only at the cost of having to assume an original paradigm utterly unlike anything otherwise reconstructible fOl' the lE verbal system.

§54. The view, originally put forth by Wagner, e-Verba, 50ff., that Hittiteverbs like teMi 'Iput' « *daiMe) , 3 sg. diii, 3 pI. tiyanzi support the reconstruction of an lE present type in *-ei-/-l:-, calls for special comment. Although these forms (Friedrich's class II 2 b) present many mOl'phological diffIculties, two facts make the comparison with the third weak class highly suspect. First, the inflectional pattern of the diii- type is not associated with stative meaning; other class II 2b verbs are lJaZziii- 'call', lJttwiii- 'run', iálJiii- 'tie', iál}amiii- 'sing', iápiii- 'eat one's fill' miii- 'prosper', niii- 'lead,9),

8) Schmid takes the 3 pI. in *-~iénti > lIr. *-yánti to have served as the starting point for the creation of fully thematic presents such as Ved. pú~y(di, mányate, etc.

9) This verb do es not strictly belong here, inasmuch as its "weak" stem is not *niya- but neya-.

63

Page 32: Jasanoff.1978.Stative and Middle in Indo-European

paí- 'give', parai- 'blo:,,', ( ?) sai- 'press' and zai- 'c.ross': Sec~nd a~d ~qually important, the Hittlte forms belong to the l¡,~-conJugatlOn: It lS thus difficult to see how they can be equated with the true e-presents and aorists ofGreek (E:[LtXv'~), Latin (habet) and Rittite itself (dannattezzi) , which show a clear preference for active (mi-) inflection. While dai- has a semanticálly specialized mi-conjugation doublet te- 'say', it is significant that this verb shows no sign of ever having contained a diphthong.

The -ai-j-iya- "apophony" which characterizes the dai-type is problem­atic. Several class II 2 b verbs are built from roots which show an enlargement *-i- in languages outside Anatolian; the clearest such case is is1Jai- 'tie', which belongs to the family ofVed. ásat 'tied' (*seh2-t). perf. sÍí/tya (*(se)soh2-j-e). Superficially, it would be attractive to compare the Rittite verb with the latter form: 3 sg. is1Jai would then continue a preform *sohd-ei, with -l¡,- introduced after s- under the influence of 3 pI. is1Jiyanzi (*sh2-i-) 11). The case of ishai-, however, is isolated: no other verb ofthe dai­class can be equated with an i-extended perfect elsewhere. The alternative thus suggests itself that dai-, pai-, ispai-, etc.~ and perhaps is1Jai- as welI~continue an otherwise extinct present type in which an ablauting root (*dheh¡-j*dhh¡-) was followed by an invariant i-element and the "activized" middle endings (with *'-e(i) in the 3 sg.; cf. §47 and § 38, note 33). This analysis need not be insisted on here; it should be clear that whatever the correct interpretation of the Rittite forms, their potential significance for the prehistory of the third weak class is slight I2

).

§55. In Lg.39, 265---6 (1963), Cowgill has suggested that both the Germanic third weak class and the Balto-Slavic stative presents in *-f­continue a single thematic type in *-3-jejo- < *-hrf,ejo-, representing the formal extension of an originally athematic suffix *-eh1-j*-hr by *-jejo­(Cowgill writes *-eE-, *-E-, *E-jejo-). Insofar as it relates to Germanic, this hypothesis is essentially equivalent to W. Bennett's reconstruction (Lg. 38, 135--41 [1962]) of the class III suffix as *-3jejo-. More recently, R. Rock has proposed a variant explanation which differs from Bennett's in sup­posing that the phonologically regular treatment of the suffix-form *-310-is preserved in Northern West Germanic *habjo and *habjanp, rather

ID) The present ofthis verb was probably fOl'med by adding the *-i ofthe hic et nunc to the root aorist *dhehrt (= Hitt. tet, Ved. ádhat).

11) Alternatively, we might reconstruct the underlying root as *sh2ehx-' and thus account fOl' i8fJ,- directly.

12) Note that under any analysis of the Hittite facts, forms like 1 sg. teMi and 2 sg. diiitti are best taken as reflecting an analogical extension of the strong stem diil:- from the 3 sg., synchronically reanalyzed as /dwi-i/. Clearly important are the old (Middle Hittite) 1 pI. forms l]alziwen and píwen, which show that the weak stem of these verbs originally ended in -i- rather than -iya-. The thematic appearance of 3 pI. tiyanzi is thus deceptive: historicaIly, the correct segmentation is /di-anzi/.

64

,.... I

than in Go. haba and haband (Kachru et al. (eds.), Issues in Ling1¿istics, pp. 332~3). For our present purposes the views of aIl three scholars may be discussed together.

The Bennet-Cowgill~Hock theory is superior to other recent explana­tions of the Germanic forms, such as those ofWagner and Schmid, in taking as its starting point a completely regular thematic paradigm free of apophonic anomalifils. In my view, however, it is very doubtful that the putative phonologiéal development of *-3jejo- to Gmc. *-ai-j*-ja-, oi *-ai-j *-a-, can be upheld. Common to all versions of the theory is the supposition that, contrary to the usual pattern in internal syllables (seen, e. g., in Go. dauhta1' 'daughter' < *dhugat¿r) , a medial laryngeal was retained in Germanic when it prece(led a *-j- which was lost before *-i-: the 3 sg. in *-aiP is thus explained by Bennett and Rock as the regular outcome of *-3jeti, which became fITst *-3jiP(i), whence *-3iP(i) and *-aiP. In a form such as the 3 pI., on the other hand, pre-Gmc. *-j- would at the outset have been preserved before *-a-. Here Rock posits a development of *-310nti to *-3janp(i), whence, after loss of medial *-3-, *-janp (cf. OS hebbiad, OE habbaP); Bennett supposes instead that *-3janp(i) yielded *-anp (cf. Go. haband, 01 hafa) , presumably by way of *-3anp( i). The crucial, and shared, assumption is thus that *-3- remained distinct from *-a- until well into the history ofGermanic prqper. For Rock this hypothesis permits an explana­tion of the syncope of *hab'Jjanp(i) to *habjanp at a date later than the change of *habajiP( i) to *hab3iP( i); for Bennett the continued existence of *-3- as an autonomous reduced vowel makes intelligible the contraction of *hab3anp(i) to *habanp (*habaanp(i), with *-a- < *-3-, would probably have yielded Gmc. *habi5np.) 13).

These arguments seem more ingenious than convincing. The loss of *3 in medial syIlables is a phonological feature which Germanic shares with Balto-Slavic, Iranian and Armenian; it is unlikely that this development occurred completely independently in all four branches 14). Direct evidence against the Bennett-Cowgill~Rock theory is provided by the class VII (reduplicating) strong verb Go. a1jan, ORG erien 'plow', a reflex of lE *h2erh3- (cf. Gk. &pOTpOV, etc.). Under Bennett's assumptions, the dialectal lE present *h2erha-jejo-, apparent reflexes of which are found in Celtic (011' 3 sg. airid), Balto-Slavic (Lith. 3 p. aria, OCS 3 sg. or'Jet'b), Greek (&p6w) and perhaps Italic (Lat. aro, -are < *arCv-ji5? ), would have been expected to yield 3 sg. *a1'3jip(i) < *a1'3ip(i) < *amiPin Germanic; in fact, however, the attested forms point unambiguously to CGin. *a1'ip (remade to *a1jiP in

13) Needless tosay, it is ofno importance to the argument whether the alleged developments took place before 01' after the Germanic consonant shift. We could as easily write *7cap~ieti for *hab<JjiP(i) and *kap~ionti for *hab~janp(i).

14) Indeed, it would probably be simplest to suppose that mediallaryngeals were never vocalized in these dialects.at aH.

65

Page 33: Jasanoff.1978.Stative and Middle in Indo-European

------~--------- ----------- --~---------

Gothic, like nasjip < *nasip), implying a development *-'Jjeti > *-jiP(i) > *_ipf5).

§ 56. Apart from p~onolo.gical considerations, *-'Jjejo- (*-h1je,lo-) seems an unlikely reconstructlOn oi the class In suffix on morphologlOal grounds. Oowgill's interpretation ofthe Germanic forms, as we have seen, is based on the assumption that Indo-European originally had a class of athematic stative presents in which full-grade forms in *-ehr ( > *-e-) alternated with zero-grade forms in *-hr (> *-'J-). It is significant, however, that in no lE language does the stative suffix *-e- appear to participate in apophonic alternations of any kind. ThiR iR most strikingJy the case in Greek, where the intransitive aorist in -1/- which Oowgill views as an old imperfect, has an invariant stem: the 1 pl. and 2 pl. end not in *-E(J.EV, *-ETO: but in -'f¡(J.O:V, -'(¡TO:, while the 3 pI. in -EV probably represents underlying *-·IJVT. The e-aorist of Balto-Slavic, although clearly a remodeled category, is similarly free of paradigmatic ablaut, as are the stative presents of Rittite and Italic. By the same token, those lE languages which have secondarily created verbal adjectives in *-to- and verbal abstracts in *-ti- from stative stems generally show *-e-to- and *-e-ti-, rather than *-'J-to- and *-'J-ti- in such forms. Thus, Balto-Slavic has both adjectives in *-eta- (Lith. -elas) and infinitives in *-íft¡: (Lith. -eti, OOS -éti) ; a formation of the same type in Germanic is Go. faheJ)s 'joy'. (On the other hand, the recent attempt by Bammesberger and Oowgill, Lg.45, 534-5, to explain the Germanic participles *habda-, *sagda- etc, (> OS gihabd, gisagd) and the abstract noun *hugdi- 'thought' (> Go. gah1lgds, OS gilmgd, etc.) on the basis ofpreformsin *-P-'J-tó, *-k-'J-tó, etc. is incompatible with the phonological history of *dnhte1'­'daughter' < *dh1lg~té1'-; the latter form shows that the application of Verner's Law in Germanic must ha,ve been later than the loss of medial schwa. See further § 77) 16).

In Italic the deverbative type in *-e-to- survives in substantivized forms like Lat. aceturn < aCMe, 1'1lbeta < 1'nbe1'e, olet1lrn < ole1'e, etc., an observation for which 1 am indebted to Alan N ussbaum. The synchronically regular participles to e-statives, however, have a suffix which appears in Latin as -ito- (cí. habitus, tacitus, placitus); that -i- in these forms reflects *-e- rather than *-a- < *-'J- is shown by Umbr, ta<;ez (written tases in the Roman

15) To be sure, it would be possible to salvage Cowgill's explanation by supposing that *-;¡- was regularly lost next to *-i- in Germanic (hence a1jan) and restored in "i}. verbs" of the haban type. But this would only add to the complexity of the axgument,

16) We could, of course, make the ad hoe assumption that *-kd~ yielded *-ht- in Germanic, rather than the expected *-gd-; this is Bammesberger and Cowgill's solution. Note, incidentally, that the evidence for a class In present from the root *hug- is restricted to the preterite *hugaidon ( = hogeta); neither a Gmc. 3 sg. *hugaifJ nor an lE *fcuk-~- has left any direct trace in our data (cf. §77).

66

T alphabet), which corresponds exactly to Lat. tacitus. The suffix *-eto- is also utilized in Italic to form the passive participle of iterative-causatives in ~-ejejo- (cí. Lat. rnonitus < nwneo, Umbr. V1lfetes beside Lat. uou.ei5), and it lS no doubt here that its origin is to be sought. Prior to the loss of Italic interv?calic *-j~, participles like *rnonetos were evidently created by replacmg the *-Jejo- ofthe present stem by *-to- on the model ofverbs like *kapjo: *kaptos, *kantajo: *kantatos, etc. After the formal merger of the pre-Italic present formations in *-ejejo- and *-e(jejo )-, the new type in *-eto­was free to spread to statives as well as causatives, whence the extension of thepattern *rnone- : *rnonetosto *habe-: *habetos, *take- : *taketos, etc, There is thus no clear evidence in Italic for an ablauting stative morpheme *-e- : *-'J-; if such an alternation was present at an early stage of Germanic it is impossible to verify this by direct comparison with forms elsewhere.

§ 57. A~ seen in § 52, the invariant -e- ofthe stative paradigm in Old High German lS probably best attributed to the analogical influence of verbs of ~he *sa:lbon class. It is far from olear, however, which ofthe two remaining mflectlOnal ~atterns, that of Old Saxon and Old English (*-ai-j*-ja-) 01'

that of GOthlO and Old Icelandic (*-ai-j-a-) is the more original 1\ While

curre~t opinion has tended to favor the former (so also, beside vVagner, Schmld, Rock and others; M. Guxman, Smv. gmrn. gerrn.jaz., 385-91), the antiquity of the Gothic-Scandinavian type has had recent defenders as well. Th~ ~ost obvious ~dvantage of as~uming a Germanic 1 sg. *habjo is the posslblhty of comparmg a form of thlS structure directly with the 1 sg. in *-joofthe Balto-Slavic presents in *-r-. Yet the value ofthis comparison is vitiated by the uncertain historical position of the latter forms: as we shall verify in the foUowing chapter, attempts to explain the origin ofthe Balto-Slavic statives have been fuUy as varied, and fuUy as inconclusive, as the theories which have been put forth to explain the third weak class.

Speaking against the Oommon Germanic status offorms like *habjois the fact that Old Saxon and Old English, the two languages which most clearly attest the stem *habja-, are precisely the languages in which the third class, as a category on the verge of total extinction, is most likely to have been con taminated with other ver balformations. In the case of the class In ver b 'live' sorne disruption of the inherited paradigm must be assumed in any event; we have seen aboye that forms like OS3sg. libad, lebot and OE liofap point to the straightforward replacement of class nI by class n inflection. In the discussion that follows a similar interpretation will be offered for the stem-forms in *-ja- of 'have', 'say' and 'live'. In particular,

17) Theoretically, of course, both could be equally old; it is also thinkable that the original type was characterÍzed by a three-way altel'llation *habai-f*haba-f*habja-. In fact there is little reason to favor either of these altel'llatives.

67

Page 34: Jasanoff.1978.Stative and Middle in Indo-European

we shall attempt to show that (*habja-, *sagja- and *libja- belong properly not to class In but to classI, and that the Northern West Germanic type *habji5, -ais, -aip is an innovation of post-Germanic date.

§ 58. Of the three ver bs which show third class inflection with ja-forms, the only one for which a stem in *-ja- appears certain to be of Oommon Germanic antiquity is 'say', which unambiguously attests *sagja- beside *sagai- in both North and West Germanic. (The corresponding verb is not found in Gothic.) It is probablynot entirely coincidental that *sagja- is also the only class In stem in *-ja- with a clear etymology outside Germanic: the Balto-Slavic present in *-í- represented by Lith. saloyti 'say' and Serb. Oh. SI. sociti 'indicare' 18) points to an original iterative-causative * sok¡'éjefo­« *sek!'- 'say'; cf. Gk. ~vvEm:, Lat. inseque) , from which *sagja- can be derived with equal ease 19). An equation of * sagja- with ES * salcí-, however, would imply the original presence in Germanic of two paradigms for 'say' , one corresponding to that ofaclassI weakverb (*sagji5, *sagis, *sagip, etc.), and the other to that of a classIII verb (2 sg. *sagais, 3 sg. *sagaip; we shall argue below for 1 sg. *sagO). This situation would clearly have been conducive to the creation of mixed paradigms in the individual Germanic dialects, and as the actual distribution of*sagai-, *saga-, *sagi- and *sagja­in North and West Germanic confirms, such secondary realignments of class I and class nI forms have in fact occurred.

The only forms of 'say' in Old Icelandic which belong unambiguously to classnI are 1 sg. segi and 2, 3 sg. segir; of these segi1', with analogical i­umlaut, ultimately continues OGmc. *sagais and *sagaip, while segi is an obvious inner-Norse creation on the basis of segi1' (segi = segir minus -1'; cf. § 51). The remaining present forms resemble those of a class I verb, save that Old Icelandic has completely eliminated the expected gemination of -g- before -j- (cf. however OSw. s<*3ggia beside sighia < segja) 20). It is noteworthy that evidence for an apparent *sagjan is not restricted to positions in the paradigm where reflexes of *sagja- are attested in Old Saxon and Old English, but is furnished also by the archaic 2, 3 sg. segr (= OSw. S<*3gh~1'; 1 sg. seg = segr minus -r) < *sagis, *sagip and impv. 2 sg. seg < *sagi < *sagje. These forms, which invite direct comparison with OS sagis, sagid (O) and OHG seg'is(t), segit, testify to the original presence of a complete classI paradigm for 'say' in Scandinavian.

The apparent reflexes of *sagis and *sagip in West Germanic have traditionally been explained as analogical creations on the basis ofja-forms like 1 sg. *sagji5and 3pl. *sagjanp (so, e. g., Flasdieck, 128). This interpreta-

18) The sense of ORuss. sociti 'aufsuchen' is perhaps due to semantic interference with *selc!l- 'follow'; cf. Fraenkel, Ll:t. Etym. Wb.,757.

19) On the development of *-eje- > *-i- in Slavic see § 86. The suggestion that *sagjan is an iterative-causative was first made to me by Hollifield in '1971.

20) Cf. Flasdieck, 121.

68

tion, however, is difficult to reconcile with the occurrence of segis(t) and segit in Old High German, where 1 sg. *seggu and 3p1. *seggent are unknown 21

); given the distribution of class I forms in North Germanic, it would be simpler to regard the pairs * sagis '" * sagais and * sagip '" * sagaip as doublets ofOommon Germanic date. Also favoring such a conclusion is the unique Old High German 1 sg. form sagu (Tatian, scribe y), which, pace Flasdieck (529), is unlikely to be a back-formation from seg1:s(t) , segit; preferable would be the assumption that, parallel to coexisting *sagis'" *sagais and *sagip"'*sagaip, Oommon Germanic had both a classI 1 sg. *sagji5 and a corresponding class nI *sagi5. Another possible reflex of *sagi5 is OE S<*330, the normal 1 sg. form in the Northumbrian dialecto The tendency ofNorthumbrian to eliminate gemination in the 1 sg. of all verbs in *-jan, however, makes it safer to regard this as a late and purely local replacement of *sec3u < *sagji5.

§ 59. Although the inflection of 'have' bears o bvious resemblances to that of 'say', the morphologicaljlifferences between the two verbs have been insufficiently emphasized by--authors who have sought to establish a paradigm *habji5, *habais, *habaip for·Oommon Germanic. In Old Icelandic the forms which supposedly establish the former presence of a stem *habja­in Scandinavian, namely 1 sg. hej and 2, 3 sg. hejr (= OSw. hav~1', with analogical absence of i-umlaut; both < *habis), are in fact indistinguish­able from the corl'esponding forms of a strong verb; descl'iptively, the plural forms hqjum, hajio, haja can be assigned to the strong conjugation 01'

to the regular class In (vaka) type, but not to a paradigm like that of segja. Of special interest is the 2 sg. impv. haj, which recalls strong imperatives like graj < graja and skjót < skjóta (contrast tel < telja, seg < segja and vaki < vaka). It is not probable that this form, occupying a naturally conservative position in the paradigm, could have been "crossed" from an inherited *haji (like valci) and *hej (lilm tel and seg). Rather, the conclusion suggests itself that 'have', like 'say' originally had two presents in Scandinavian, one, *habi5, *-ais, *-aip, cOl'responding to that of a class III verb ofthe regular type, and the othel', *habi5, *-is, *-ip, formally identical with that of a strong verb of classIV.

The Germanic, rathel' than purely Scandinavian character of2 sg. *habis an:d 3 sg. *habip is suggested both by OS habis, habid (O), with analogical absence ofumlaut as in sagis, sagid, and by OHG hebis(t), hebit, which, like segis(t) and segit, can only with difficulty be regarded as analogical cl'eations on the basis of unattested *hebbu, *hebbent, etc. Although both Old Saxon and Old English attest unambiguousja-forms in the 1 sg. and elsewhere, it is not impossible that the replacement of, e. g., *habi5 (originally proper to both the stl'ong and class nI weak pal'adigms) by

21) On the distribution of OHG stem forms see Brinkmann, Sprachwandel, 96ff. and K6geI, PBB 9, 518 (1883).

69

Page 35: Jasanoff.1978.Stative and Middle in Indo-European

*habji5 was triggered by the generalization in these languages of *sagji5 (originally class 1) at the expense of *sagi5 (originally class HI). An important indication that the stem *habja- may be secondary in Northern West Germanic is the consistent absence of umlaut in the ja-forms of 'have' in Old English (cf. hoobbe, habbap, habban, etc. beside sec3e, sec3ap, sec3an) and its sporadic absence in Old Saxon (cf. habbiu, habbiad, habbian (M), habbiu (C), beside invariant seggiu, seggiad, seggian) 22). Direct evidence for a West Germanic 1 sg. *habi5is found in ORG habu, attested twice in the same source (Tatian, scribe y) which preserves 1 sg. sagu, and perhaps in OE (Northumbrian) hafo.

The obvious wealmess of assuming the former existen ce of an indepen­dent paradigm *habi5, *-is, *-ip in Germanic is that a comparable present is found nowhere else in Indo-European. (It is not impossible, of course~ that the thematic inflection of *hab- simply represents the replacement of an earlier athematic type like that of Alb. karn '1 have < *kabhmi 01' *kapmi.) Whatever the genesis of the strong forms of 'have', however, the hypothesis of two distinct presents accords well with the facts in the case of 'say', where the reconstruction of an independent *sagjan is supported by Lith. sak~ti and Serh. Ch. ~I. s~éiti. And while the expedient of assuming an earher twofold paradlgm lS perhaps somewhat inelegant, it is at least significant that fewer ad hoc assumptions are needed to explain the special peculiarities of 'have' and 'say' under the analysis suggested aboye than under the competing view that both verbs, like the membership ofthe third weak class as a whole, were originally characterized by the alternation of *-ai- and *-ja- within a single paradigm.

§ 60. Forms pointing to a Gmc. 1 sg. *libji5, 3 pI. *libjanp, etc., like those pointing to *habji5, habjanp, etc., are confined to Old Saxon and Old English. EIsewhere in Germanic the inflection of 'live' is perfectly regular, save for the infrequent appearance in Old Righ German of 2 sg. libis and 3 sg. libit; OSw. 2, 3 sg. livl¿1' beside l'iv'ir (as ifto 01 *lifr beside lifir) is almost surely to be attributed to a secondal'y extension of the inflectional pattern of 'have' in East N 01' se 23). Pro bably the only certain conclusion that can be drawn from the West Germanic fOl'ms of 'live' is that this word has everywhere shown a pronounced tendency to acquire new paradigms of a synchronically deviant type. It is at least possible that ORG libit was coined on the model of hebit and segit; favoring an early morphological assimilation of 'live' to 'have' and 'say' might have been the fact that all

22) See Holthausen, Alts. El.2, 164. This differeilCe between 'have-' and 'say' has never been adequat~ly e.xplain.ed. A different view is taken by J. Dishington, who in a personal commumcatlOn pomts out that the absence of umlaut in h;;ebbe, habbi7¿ may simply reflect a desITe by Old English and Old Saxon speaker s to distinguish these forms from hebbe, hebbiu '1 lift'.

23) For other such forms (dugh~r, pol~r, etc.) see Flasdieck, lOG, cit.

70

T

three verbs, together with *hugjan 'think, ponder' (class 1), originally had irregular weak preterites with the phonologically marked clusters *-bd- and *-gd- (cf. OS hab~a, sagda, ho.gda, libd([, ORG hapta, hocta) 24). Although it is le~s e.asy toenvlSag~ ~ ~peClfic channel for the reJ?lacement of, e. g., 3 sg. *ltba~p by class H *Mop m Old Saxon and Old Enghsh (cf. § 49), it is evident that the mixed inflection of 'live' in thése languages casts considerable doubt on the view that forms like 1 sg. OS libbút, OE libbe and F----3 pI. OE libbap are genuine inheritances.

§61. Thus, confused and ambiguous as the histories of 'have', 'say' and 'live' undoubtedly are, it is at the very least clear that these verbs do not constitute real evidence against the view that the Common Germanic

__ class IH paradigm was of the Gothic-Scandinavian type, i. e., characterized by thé regular alternation of stem-forms in *-ai- and *-a-. On the other hand, it is notable that scholars who have taken the opposite position and claimed Germanic antiquity for the type *habji5, *-ais, *-aip have without exception been unable to adduce a plausible motivation for the replace­ment of*-ai-j*-ja- by *-ai-j*-a- in N orth and East Germanic. As mentioned in § 53, the usual explanation for the creation offorms like Go. haba, haband, etc. and 01 vqknm, vaka, etc. has been to assume that Gothic and Scandinavian systematically eliminated *-j- from the class nI paradigm under the influence ofj-Iess forms like CGmc. 3 sg. *habaip, *wakaip, etc. But this will not explain the special behavior of 'have' and 'say' in Scandinavian, which differ both from other class nI verbs and from each other; nor does it seem natural to suppose that if forms in *-ja- were felt to be anomalous they would not simply have been replaced by forms with *-ai- generalized from the 2 sg., 3 sg. and 2 pI., as happened independently in Old Righ German and, in part, Old English (cf. § 51).

On balance, therefore, the available evidence favors a reconstruction of the present paradigm ofthe third weak class as sg. *habi5, *habais, *habaip, pI. *habam, *habaip, *habanp, inf. *haban. In the following sections we shall attempt to discover how this pattern originated.

§ 62. As §§ 52--6 have shown, attempts to copnect the paradigm of the third weak class directly to the stative suffix *-13- 01' to one of its presumed variants, have not in general been successfuI. It is important to note, therefore, that forms such as the class nI 3 sg. in *-aip need not from a diachronic point ofview reflect,the addition of a desinence *-p to a proper suffix *-ai-, but may instead owe their origin to the historically redundant suffixation of *-p to a well-formed 3 sg. in *-ai, already provided with a desinence. Morphological renewals of this kind are not infrequent in lE languages. We have already (§ 47) called attention to the replacement of Ved. áduha, *ásaya by "thematic" ádnhat, ásayat, and to the similar

24) The status of these forms is discussed in § 77 below,

71

Page 36: Jasanoff.1978.Stative and Middle in Indo-European

------- --- ---------------

replacement of *1I-idé by *1I-idét in the prehistory of the lE thematic aoristo In Hittite the 3 sg. in -ai ofthematic b-i~verbs can be remade to -aizzi, thus falling together with the normal 3 sg. of (mi-conjugation) verbs in *-ehé:ejo­(= Brugmann's *-i'ijejo-): typical instances are tarni'iizzi beside tami'ii 'leaves', and iSkalli'ii beside iSkalli'iizzi 'rips up', the latter found in the earlier and later recensions of the Laws, respectively25). Note also the type parsiya--+parsiyazi discussed by Watkins, ldg. G1·. III. 1, 102. Cases like these are scarcely to be separated from such developments as the replacement of the Vedic 3 pI. middle ending -ra by -ran 01'

-rata, the replacement of the inherited 2sg. middle in *-tha « *thze) by Ved. -thi'il¡" or the renewal ofthe 1 sg. middle ending *-r fJ.laas -fJ.~:-1 in Greek.

In Germanic two 3 sg. forms in *-aijJ for which an análysis of this kind suggests itself are *rí:raijJ (val'. *rairaijJ) 'trembles' (cf. Go. reiran < *1'i'-, OE ri'irian<*rai-) and *bibaijJ (val'. *bebaijJ, *b'ibaip) 'id.' (cf. OHG biben, beben, OI bija, OSw bejjwa, OE bifian). It is usual to compare Go. 3 sg. reiraijJ with Skt. leldyati 'swings, is unsteady'. The form leláyati, however, is found no earlier than the Brahmanas, and is itself the replacement of an earlier 3 sg. leldya (MS), a wholly anomalous creation which resembles an intensive in reduplication but a perfect in inflection. (Compare perhaps Hitt. asasi 'settles' beside esari 'sits'.) Similarly , the starting point for the creation of Gmc. *bibaijJwas probably a perfect form similar to Ved. bibhdya 'is afraid' (the lE prototype was presumably *bhebhóihxe)26) rather than a secondary present like, Ved. bibhéti; the long vowel of the variant *b'íbaijJ suggests that this verb too may have had byforms with intensive reduplication 27).

The process by which forms like 3 sg. *reiróihxe and *bhe(i)bhóihxe were replaced in Germanic by *r'íraijJ and *bltbaijJ is not, to be sure, entirely clear. The possibility cannot be excluded that the endings of the (athematic) active were substituted at an early period for those of the perfect, in which case *bibaijJ would represent a form precisely analogous to bibhéti. It is at least as simple, however, to suppose that *r'íraijJ and *bltbaijJ were created in the same way as leldyati, i. e., by the direct suffixation of *-jJ(i) or *-ti to an already complete 3 sg. formo Once established in the verbal system of Common Germanic, *r'íraijJ and bltbaijJ were treated as normal classIII forms and equipped with full classIII paradigms 28

).

25) This interpretation of tarnaizzi, etc. was first suggested by Watkins, Id(¡. Gr.III. 1, 102.

26) The -i- ofthe first syIlable ofbibaip, of course, would have been phonologicaIly regular before the *-bt(i)- of the plural forms.

27) This would accord weIl with the semantics of *bibaip. That >!<bibaip and bibhéti represent independent innovations was seen long ago by Wackernagel, Kl. Schr., 305-9 [1907]. 1 am indebted to Hollifield for calling my attention to leláya.

28) PhonologicaIly, it would then be most natural to as sume that *-oihxe, 01' its regular development *-aije, was extended to *-oihxeti (*-aijepi); this gave flrst *-aijijJi, and subsequently, with the loss of *-j- before *-i-, *-aijJ.

72

§ 63. The possibility of an analysis of this kind for the two verbs just discussed, taken together with the results of the preceding chapter, suggests a new approach to the problem ofthe third weak class as a whole. In our study of the class III and IV presents of Tocharian we saw that although the roots from which such presents are derived frequently underlie e-statives elsewhere in Indo-European, the Tocharian forms themselves ultimately represent transformations of the lE middle root aoristo It is thus not possible to refer Toch. B o8otiir 'turns dry' and Lat. 1iYere, 01' Toch: B wokotiir 'splits, opens (intr.)' and Gk. (f)ay"~vaL to a single original paradigm; the existence of these and similar pairs simply reflects the fact that many roots which formed a stative stem in *-e- in late Indo­European denoted the punctual entry of a subject into a state, and thus also served' as the basis for a middle root aorist, together with suoh formations as, e. g., a stative perfect and an "intransitive" present in *-jejo­(cf. §§ 3,9). In principIe, there is no reason why the relationship ofthe third weak class to categories su eh as the Latin statives in -e- 01' the Greek intransitive aorists in -'r¡- cannot be of this kind also: correspondences like Gmc. *lubaijJ: Lat. lubet, OHG dagen: Lat. tacere, and OHG ji1'-monen: Gk. (J.av·~vaL can as easily be explained by assuming that the lE prototype of Gmc. *-ai- and the stative suffix *-e- tended to be associated with the same roots, as by supposing that Gmc. *-ai-, Lat. -e- and Gk. -'r¡- were originally mere variants of one and the same underlying morpheme 29

).

In my view, an earlier version of which was set forth in Lg.49, 850ff. (1973), the class III 3 sg. in *-aijJ is ultimately to be analyzed as a 3 sg. middle in *-ai < *-oi, secondarily reinterpreted as part of the verbal stem and suffixed by the 3 sg. active ending *-jJ. In the following sections we shall attempt to identify the lE middle formations which can plausibly be viewed as having undergone this renewal, and to show how the replacement of *-ai by *-aijJ in the 3 sg. can ultimately have led to the establishment of the class III paradigm in its Common Germanic formo

§ 64. The Gothic and West Germanic preterito-present *dugan, 3 sg. *daug 'be fit, be strong, taugen' (cf. Go. dugan, daug, OHG tugan, toug, OE dU3an, dea3) is not found in Scandinavian; its place is taken in Old Icelandic by the class III weak verb duga, 2, 3 sg. dugir « *dugai8, *-aip), which in part continues the sense of the old preterito-present, but in part has the non-stative meaning 'help, render assistance'. The etymology of these forms has been discussed by Meid, Da8 germ. Praet., 24-5, who has established their connection with the lE root *dheugh- 'give forth as a benefit, Ertrag produzieren'.

Outside Germanic, primary verbs from this root are found in Greek ('l'e:ÚXw, 'l'Unávw) and Indic (3 sg. duhé 'gives milk; milks'). Neither Gk. 'l'e:úxw nor 'l'U'YXávw is likely to continue an inherited present: 'l'e:úxw, which

29) This was seen very clearly by Kurylowicz, Infl. Oat., 82.

73

Page 37: Jasanoff.1978.Stative and Middle in Indo-European

resembles secondary thematic forms such as Aebtw, Tép1tw, Tépcro[J.cu, etc., lacks an exact counterpart in Vedic 01' Germanic; Tunávw, like other "double nasal" presents in Greek, is probablrc a late formation on the basis ofthe corresponding thematic aorist (huxov) 0). The archaic morphology of Ved. duhé on the other hand (3 sg. mido -é for -té, 3 pl. duhTé beside duhaté, etc.), identifies it as a form ofIE antiquity; the associated active dógdhi ís later, being confined in the Rigveda to a single.inst~nce of3 sg. ádho.k (4'.1? 7), and to fourteen occ~ren.ce~,of 3 p~. ~uhánt~, ,WhICh m~y _owe th81r orI~~n to the influence of the actIve partIClple d1lhant- ( = d1lhana-; cf. § 39) ). For Indo-European we may reconstruct an athematic 3 sg. *dhughó(T) ~s the immediate precursor of duhé; su eh a form may also haveplayed a role m the creation ofthe pre-Greek thematíc aorist *dhughét (> (~)Tuxe; cf. §§47, 100). Together with the perfect *(dhe)dhó1lghe (> Gmc. *da1lg), *dhughó(r) is the only 3 sg. form which may be assumed for the verbal system of the earliest pre-Germanic. § 65. The form *d1lgC(ip is most simply regarded asthe e~~ensio:r: in *-p ~f earlier *d1lgai, a 3 sg. middle comparable to Ved. d1lhe l. In ltself thlS analysis need not have a bearing on the interpretation of the third class as a whole: the complete formal overlap of *d'ugan, *-aip with class In verbs such as, e. g., *wakan, *-aip could in principIe be explained by supposing that *dugan was morphologically assimilated to *wakan after the seconda­ry coincidence ofthe two verbs in the 3 sg. On the other hand, we have seen in §§ 52----6 that traditional attempts to derive the inflection of *wakan and other statives from an earlier type in *-e- have proved unworkable on both phonological and morphological grounds. It is thus of cons~derable int~rest that the paradigm of the third weak class can be explamed as a dlrect transformation of the athematic middle type to which 3 sg. * d1lgCti , like duhé, must originally have belonged.

Letus assume that at a stage ofpre-Germanicprior to the loss ofthe non­passive middle the ancestor of olass nI *dugan had a present of the following type 33) :

30) The semantic difference between Teúxw and TUYXávw is largely to be explained from the "reciproca!" sense of the root *dheugh-, which dou btless mean t 'receive (as a product)' as well as 'produce' in the parent language. Cf. Benveniste, Probo de ling. gén., 315-26 [1951).

31) The other form which must have played a role in the creation of d1thánti was the "active" imperfect ádt¿hat ( = ádt¿ha + t); ád1/,hat and d'z¿hánti·in turn gave rise to occasional class VI transfer forrns in the later language.

32) That the primary rniddle endings in Germanic were characterized by .*-i rather than *-r is shown by the Gothic passive endings -ada, -aza and -anda, WhICh continue *-ada1:, *-azai and *-anda,i. .

33) 1 have, of course, no way of establishing a relative chronology of the morphological changes to be discussed in this section with respect to Grimrn's Law and Verner's Law. The argument would in no way be affected ifwe wrote *dhugh­for *dug-, *-'(ftoi fOl' *-ttndai, etc.

74

sg.l *dugai (cf. Ved. -él 2 *d1lhtai (cf. Hitt. -(t)ta, Ved. secondary -tMilp) 3 *d1lgai (= Ved. d1thé)

pl. 1 *dugm- 34) 2 *d1tgd- 34) 3 *d1lg1mdai or *d1tgmi (cf. Ved. d1thaté, d1lhré)

Such a paradigm could not have survíved indefmitely: the complete absence ofrecognizable deponents in Common Germanic makes it virtually certain that the above forms would eventually have been restructurecl as actives 01' eliminated entirely from the language. We have suggested that the former was the course acloptecl: the obsolescent 3 sg. ending *-ai was reinterpretecl .as a' bare stem-vowel ancl enlargecl by the *-p of the 3 sg. active 3,,). In view of the absolute regularity with whích the 3 sg. of the Germanic present agrees in preclesinential vocalism with the 2 sg. ancl2 pl., it ís very probable that thís renewal woulcl in turn have entailecl the creation of a parallel 2 sg. *d1lgais ancl 2pl. *dugailJ « *-pe).

How the 1 sg., 1 pI. and 3 pI. might have acquirecl active enclings is less olear. Ifthereplacement of*-ai by *-a'ip in the 3 sg. hacl been the initíal step in the "activization" of *d1lgan, we should perhaps have expectecl that the new stem *d1lgai- woulcl be generalized throughout the paracligm, giving rise to forms like 1 sg. *dugai, 1 pI. *d'ugaim, 3 pl. *dugainp (cf. class n 1 sg. *salbo, 1 pI. *salbom, 3 pl. *salbonp, with *-0- as in 3 sg. *salbojJ). But it is equally possible that the creation of *dugaijJ occurrecl at a comparatively late stage in the elimination of the miclclle paracligm. In the 1 sg., the inherited form *d1lgai was homonymous with the 3 sg.; here, as Hollifielcl has pointed out to me (personal cómmunication; cf. § 78), *-ai coulcl easily have been directly replaced by the active encling *-0, which characterized the great majority of 1 sg. presents in Germanic. Similarly, the substitutíon of 3 pI. *duganp for *d1tgrai (01' *dug1tndai) may have been triggered by the presence of an archaic nt-participle *dugand- 36); the P?si~ion ?f *d1lg~np would then be precisely analogous to that of Ved. duhant~ besIcle d1thant­(cf. § 64) ancl, mutatismutand1:s, Toch. A 3pl. Titweñc beside ptcp.1'itwant (cf. § 39). If active forms such as 1 sg. *dugo and 3 pl. *duganp already belongecl to the paracligm of *d1tgan when *dugai was replacecl by the new 3 sg. *d1lgailJ, it would not be clifficult to unclerstancl why the analogícal spread of the stem *dugai- was restrícted to the 2 sg. and 2 pI.37

).

34) It may be supposed that Germanic once had plural endings ofthe type seen in Ved. pI. 1 -l1whé, 2 -dhvé, Gk, -¡J.eeot, -creE. Since we cannot be sure of the forrns of these endings in Indo-European 1 have left them unspecified.

35) Pre-Grnc. *-ai hadno doubt already been replaced by *-dai « *-toi) in sorne of its original functions.

36) On the status of *d1lgand- see § 75. 37) Note the typologically similar spread ofthernatization in the frrst person of

the Slavic aorist: 1 sg. delax'b < *-som, duo delaxove < *-soye, pI. delaxom'b < *-somos, but 3 sg. dela < *-s(t), duo delaste < *-ste, pI. dllaff~ < *-S'(ft.

75

Page 38: Jasanoff.1978.Stative and Middle in Indo-European

-As we shaIl see below, further light is shed on the remodeling of the

athematic middle in Germanic by the history of the weak preterite. The aboye account, however, has been sufficient to demonstrate the theoretical possibility that the *-ai-j*-a- stem vowel characteristic of the third weak class reflects not an etymological suffix, but a morphological process whereby active desinences were adjoined to 01' substituted for their obsolete middle counterparts. In the last analysis, the practical signifi­cance ofthis possibility must depend on the extent to which class III verbs other than *dugan can be compared with semantically appropriate middle forms elsewhere in Indo-European. As it happens, pairs like Gmc. *dugaip and Ved. d~thé are by no means isolated.

§ 66. Under the analysis just proposed the paradigm of *dugan represents a transformatíon of an athematic middle type similar to that presupposed by Toch. A parkatar (III) 'goes up', B lyuketar (III) 'shines', B wolcotar (IV) 'splits, breaks' and similar forms. The two formations differ in only one respect: while it was seen in ch. 2 that the Tocharian forms are ultimately based on inherited root aorists, pre-Gmc. *dugai and Ved. duhé can most directly bereferred to an lE present *dhnghó(r). The status ofthis form will be discussed further in § 100.

By way of digression, it may be noted that originally deponent aorist presents ofthe Tocharian type are also represented in Germanic, but only as a marginal class. The clearest example is probably *f~tlgan, *-aip 'follow' (cf. ORGfolgen, OSw. sg.folger (: OIhapaxfnlger, Flasdieck, 59), further OS folgon, OE fol3ian, with the characteristic transfer from classIII to class II). In the absence of a better etymology, 1 would submit that this word is cognate with Ved. sprs- 'touch' (pres. sprsáti), which means 'reach out' in RV 4.4.2 táva bhmmása íisuyá patanty ámt sPl"sa dhrfjatá sósncéinalt, translated by Geldner 'deine Lohen fliegen schneIl, greif mutig zu, hell flammend' ; compare also the compounds nparispts- 'towering' and l"taspts­'pursuing, practicing rta-'. Although no root aorist of spis- is actually attested, the class VI present spl"sáti presupposes a thematic injunctive *sprsát and is thus probably ultimately referrable to an athematic 3 sg. form in *-e 01' *-0 (cí. §§47, 100). For Indo-European we mayreconstruct a 3 sg. aorist middle * (s)plfó 38

) ; at an early stage ofpre-Germanic such a form evidently served as the starting-point for the creation of an aorist present *p[kói, which in turn became the immediate source of *fnlgaip 39). A different present formation is reflected in the class 1 weak verb *fnlgjan (01 fylgja, OE fyl3(e)an), originally a jejo-present of the fLdvOfL<XL-type; the

3H) Whether *~o was here original or the replacement of earlier *-e is immaterial for the present argument; as we have seen already, *-e was everywhere replaced by *-0 (01' *-to) in forms which continued to function synchronically as middles.

39) The Old English preterite full eode 'followed' rests on a misinterpretation of 3sg. *fulgaiP as a compound of 3sg. *gaiP 'goes'.

76

relationship of *fnlgjan, *-ip to *f~tlgan, *-aip is precisely the same as that of Ritt. par'kiya- to Toch. A parkatar (§ 33). Outside Germanic and Indo­Iranian the root *(s)pelk- 'reach for, strive' is found also in Toch. B spíilk­'be eager, strive' (A spíiltk- with secondary -tk-; cí. also B speltke 'eagerness') 40). Gmc. *fnlgan, Ved. sPl"S- and Toch. *spíil(t)k- share the syntactic peculiarity of governing an object in the locative (= Germ. dative) case.

§ 67. CGmc. *dugan, *-aip is one of an archaic group of class III verbs built to roots which al so underlie preterito-presents (cf. Go. dang, etc.). The other mem bers of this group, which are discussed at length by W aguer, e­Verba 7ff.; are '*mnnan, *-aip (Go. munan 'intend, wish', ORG fir-monen 'despise'; cí. Go. man '1 think', etc.), *witan, *-aip (Go. witan 'observe', OE be-witian (II < III) 'id.', ORG gi-, ir-wizzen 'be capable'; cf. Go. wait '1 know', etc.), *parban, *-aip (Go. ga-parban sik 'abstain' [with 3sg. -ip erroneously for -aip, 1 COl'. 9. 25], ORG dat'ben 'lack', OS tharbon (II < III) 'id.', cí. Go. pa.rf '1 need', etc.), *kunnan, *-aip (Go. ga-lcunnan 'recognize', ORG knnnM 'id.'; cf. Go. kann '1 lmow how', etc.), perhaps also *magan, *-aip (ORG magen 'be able', OE ma3ian (II <III) 'prevail' ; cf. G<? mag '1 can', etc.)41).

The analysis of *dugaip presented aboye makes it a priori likely that the synchronically parallel forms *m~tnaip, ~witaip: *parbaip,. *lc,,!,nn(1) and *magaip reflect remodelings of a similar kmd. ThIS expectatlOn lS borne out by a considerable body of comparative evidence.

§ 68. The forms of the root *men- 'bring to mind, remember' reconstruct­ible for Indo-European include a deponent jejo-present (Ved. mányate, Glr. fL<x1vofL<XL, 011'. dO'moinÍ1tr), a deponentroot aorist (Ved. ámata, GAv. man,tíi) and a stative perfect (Gk. fLÉfLoV<X, Lat. memin'i, Go. man), the latter bemg the only "active" form traceable to the parent langu~ge \cí. § 3). In principIe it would thus be reasonable enough to seek a denvatlOn of Gmc. *mnnaip from an earlier 3 sg. middle in *-a'i < *-oi; no other attested language, however, offers any evidence of a deponent root pres~nt from which pre-Germanic could have inherited a 3 sg. *mnnai. Alternatlvely, w~ might attempt to explain *mnnai as an aorist l;lr~sent lik~ pre.-Gmc. *fnlg~~ (§ 66), but the stative meaning of *mnnan, *-a~p m the hlstoncal Germamc languages would not at first glance favor such an interr~retation.

On the other hand, it is significant that the root *men- lS one ofthe very few for which it is possible to reconstruct a perfect middle in Indo-

40) It is simplest to assume that CToch. *-ltk- was reduced to *-lk- in Toch, B. The vocalism of the Tocharian forms is problematic.

41) 1 omit the pair 01 óask 'fear': Go. ogan 'id.', since only th~ prete~'i~e of the Scandinavian verb shows classIIl forms (cf. §77). Go. 7magands unafrald ,on the

. t d * *'" other hand, perhaps does reflect an otherwlse unattes e agan, -aty .

77

d

Page 39: Jasanoff.1978.Stative and Middle in Indo-European

European: a 3 sg. *(me)mnól' (*-ói) isdirectly 01' indirectly attested in Indo­Iranian, Greek and Celtic, the only three branches of the family in which the perfect middle occurs as an autonomous category. IIr. *mamnai is attested as the perfect of *manyatai twice in the Rigveda (2 duo mamnáthe 7.93.6, 3 duo mamnCite 7.31. 7) and, significantly, no fewer than four times in Avestan (Gathic 3 duo mamanaite Y. 13.4, YAyo 3 sg. mamne P. 18, 25, ptcp. mamnanai Yt.13. 88), where the perfect middle is far less productive than in Vedic. N either Indic nor Iranian has a corresponding perfect active. In Old Irish the suffixless preterite of dO'moinethar is dO'ménair, which appears to reflect a pre-Celtic 3 sg. *memnor, with analogical palatalization of the final consonant; cf. Watkins, ldg. Gmm. lII. 1, 222, and note 44 below. Greek attests the perfect middle of *men- in the form fLÉfLv'I)'t'OCL,

showing the apparent substitution of the set zero-grade -fLV'l)- for ghonologi­cally regular -fLOC-, as in the corresponding present fLLfLv'~O')WfLOCL ). From a functional point of view, it is of interest that *(me)mnór (*-ói) is synchronically associated with apresen t in all three branches of the family; in this respect it differs characteristically from the perfect active *(me)móne, which functions as an independent lexical item in Gk. fLÉfLoVE,

Lat. meminit and Go. mano It would probably not be far off the mark to suppose that this was already the situation in late Indo-European, and that *(me)mnór was originally created to fill the structural gap caused by the seman tic divergence of the inherited perfect * (me )móne and the presen t *mn-iéjó-. Th~ assumption of an lE perfect middle *(me)mnór permits a simple

formal and functional explanation of Gmc. *munaiP. Since reduplication is ordinarily lost in the Germanic perfect system, the expected pre-Germanic reflex of this form would have been *munai 'remembers, thinks, intends', with the zero-grade *mun- (for *mn-) either generalized from anteconso­nantal position 01' regularly derived from a variant *m'ftnói (by "Lindeman's Law", cf. NTS20, 38-108 [1965]). As in Indo-Iranian, the retention of the dentalless ending *-oi in the 3 sg. of the perfect middle can presumably be attributed to the absence of a dental in the corresponding active ending *_é3

). The inflection of *(me)mnár in late Indo-European would have been identical with that of *dhughór; the same analogical

42) Such substitutions are common in Greek. They are in part to be explained from the fact that since se~ and an1:t roots commonly had antevocalic zero-grades Qf the same structure (TaR-), their anteconsona:t!tal zero-grades (T1;l- fol' anit, TRA­for set roots) tended to be equated also. TRA-, which spread in this way to anit roots, was subsequently extended to antevocalic environments also. The productiv­ity of such zero-grades is nicely illustl'ated by the presents 8v-ncrxw and (Aeolic) fL~fLvlX(crXW, which have apparently replaced earlier *6IXv(crxw(set root) and *fL~fLv(crxw (01' *fL~fLlXv(crxw; anit root), respectively.

43) In Common Greek, of course, -TOL (> -TIXL) was extended to aH middle paradigms with primary endings.

78

'r I 1

I developments may be assumed for the paradigm of Gmc. *munan, *-aiP as for that of *dugan, *-aiP.

§ 69. This analysis can be extended without difficulty to *witan, *-aiP. For the root *'l}eid- too there is reason to suppose the existence of a perfect middle in late Indo-European. The Vedic form vidé, though listed by Grassmann as a present of vid- 'find', has been shown by Renou, Val. du parj., 155ff., to bethe 3 sg. middle corresponding to véda 'knows'; although the two forms are not consistently distinguished in meaning, vidé is sometimes passive, and has a specifically medial nuance in such passages as 7.40.5 vidé hí rudró rudríyam mahitrám 'Rudra connait en effet la grandeur propre a Rudra;'. The Rigveda contains eighteen occurrences of vidé, to which may be added more than twenty instances ofthe participle vidCiná­(vídana-) , three of 2sg. vitse and two of 3pI. vidré. Even though no equivalents of these forms are quotable from Iranian, they are clearly archaic: it is instructive, e. g., to note the formulaic use ofthe phrase yathá vidé 'as one lmows, as is known', which is attested six times in the Rigveda, always in pada-final position. In both structure and meaning, moreover, vidé strongly recalls the Old Irish preterito-present ro'fitir 'knows' (1 sg. ro'fetar). The latter verb presents well-known difficulties: both the root vocalism and the unlenited -d- of 'fitir point to an immediate preform *widr' < *widri 01' *widre, while a perfect middle of the expected type (*widor) would havé yielded OIr. *'fedar. Thurneysen, OIGr., 436, has suggested that the Old Irish paradigm was remodeled on the basis of a 3 pI. form akin to Ved. vidré; alternatively, it is not impossible that an inherited *widor was remade to *widri under the influence of the 3 sg. deponent ending *-tri (cf. §46)44). Whichever analysis is ultimately preferred, it is difficult to escape the conclusion that the inflection of ro'fitÍ1' must in the last analysis be derived from an athematic middle type similar to that of vidé45

).

From a formal point of view, of course, the development of *~idór to *witaip in Germanic would exactly duplicate that of *dhughór and *(me)mnól' to *dugaip and *munaiP. The semantic link between *witan, *-aiP and the perfect middle forms ofIndic and Celtic is clearest in ORG gi­wizzen 'be capable (+ gen.)', the intransitive value of which recalls the frequent use ofVed. vidiiná- (vídana-) in the absolute sense 'wise'46). Go.

44) MW g(vyr 'Imows' points ultimately to the same preform as fiti?', cf. Meid, Ét. Celt. 13, 346--7 (1972).

45) The synchronic association of ro . fitir with the present ro· finnadar 'gets to know' recaHs that of do· ménair with do' moinethar. Nowhere in Indo-European, on the othel' hand, is the active pel'fect *J!Óide synchronicalIy associated with a presento

46) The "active" participle vidv(i1~s- is also employed absolutely, but its Germanic cognate has been specialized in a different meaning (cf. Go. weitwops 'witness').

79

Page 40: Jasanoff.1978.Stative and Middle in Indo-European

witan, -aiP 'observe, watch (+ dat.)' is similarly intransitive, but has durative, rather than stative meaning; for a possible explanation see § 72, note 51. § 70. Of the remaining class III verbs associated with preterito-presents, Gmc. *pa1'ban, *-aiP shows an aberrant o-grade ofthe root, which precludes the possibility of a direct equation with the Vedic perfect middle tiill'pé < trp- 'become sated'. It is striking, however, that * pa1'ban agrees exactly in vocalism with the Toch. A class IV present tsa1'watii1' 'is confident' < *tOl'potOl', a form tentatively analyzed aboye as continuing an lE acrostatic present *t01'p-j*te1p- (§ 38). In theory, the possibility cannot be excluded that the Germanic and Tocharian verbs represent completely independent creations, and that *parban is a denominative formation

based on *pal'bo 'necessity' (= Go. pal'ba) 01' *Par'ba- 'necessary' (= Go. pal'bs). Yet neither the semantics of *pa1'ban nor its obvious structural parallelism with *munan, *-aip, *witan, *-aiP, etc. speak in favor of such a view47

); it is simpler to suppose that the Germanic and Tocharian forms reflect a single present middle paradigm of lE antiquity. See further § 72.

Little can be verified about the prehistory of Gmc. *magan, *-aip 'be strong'; indeed, the very legitimacy of assuming such a verb for Oommon Germanic on the strength of ORG magen and OE ma3ian (class lI) is not entirely beyond question. Yet it is difficult to see how these forms could have been created on the basis of the preterito-present *mag in post­Germanic times, and correspondingly attractive to consider the possibility that, whatever their status at the lE level, an active perfect *mag « *mage) and a middle 3 sg. *magai existed side by side in the verbal system of pre-Germanic48

).

The relationship ofthe third class verb *kunnan, *-aiP to the preterito­present *kann is discussed in §73.

§ 71. The explanation just proposed for *munan, *-aiP, *witan, *-aip, * p~a1'ban, *-aiP and *magan, *-aiP depends critically on the assumption that

47) Although *-0- originally characterized only the strong forms of the inherited paradigm, the root vocalism of late lE *torpor (for expected *terpor) can be understood if we assume, as suggested in ch. 2, note 33, that presents of this type were originally "deponent", i. e., that like thematic presents, they ended in *-e, rather than *-ti, in the 3 sg. With the synchronic reinterpretation of *-e as an active ending (§45), it would have become possible for *-0 to be substituted for *-e informs with overtly middle value: in this way *bhem(r) was introduced as the new middle corresponding to 3 sg. *bhel'e, and *torpo(r) became the middle counterpart of *torpe.

Gmc. *pm'ban, *-aiP must be carefully distinguished from 01 parfa (II) 'be necessary', which represents a true denominative * parbon.

48) There is no semantic justification, however, for supposing a historie al connection between *magaip and Ved. mamahé (5 x ), which consistently appears to mean 'granted'.

80

ya

the perfect middle was already present as a category in late Indo­European. In recent years this assumption has been questioned by a growingnumber ofscholars (see, e. g.,Kurylowicz, Infl. Cat., 63), whohave taken the late extension of the perfect middle in Indo-Iranian and Greek (cf. §4) to indicate a post-IE origin for the formation as a whole. In fact, however, it is often difficult to distinguish a category which has been independently innovated in several lE languages from one which was marginally present in the proto-Ianguage and independently elaborated in the dialectal periodo The equations Av. mamne = Gk. f1.~f1.V"t)To(' = 011'. 'ménail' and Ved. vidé = 011'. fitil' constitute strongpr'imafacie evidénce for assuming at least two perfect middles for late Indo-European; to the extent that tlie account given aboye ofGmc. *munaip and *witaip provides a simpler and less al' bitrary explanation for these forms than the theories surveyed in §§ 52----6, Germanic may without circularity be said to contribute to the same conclusion.

Taking the lE status of *(me)mnól' and *Uidól' as given it is natural to suppose further that, as in Indo-Iranian, Greek and to a lesser extent Oeltic, the perfect middle underwent a period of secondary expansion in the early history of Germanic. Such a conjecture cannot be verified directly; it can be made plausible, however, by showing that a munber of third class verbs correspond etymologically to lE roots of a type which proved favorable to the creation of new perfect middles elsewhere in the family.

In the attested forms of Indo-Iranian and Greek, the perfect middle is a fully established constituent of the verbal system, and new perfect middles may be created at any time by applying the middle endings to the appropriate (originally zero-grade) form of an already existing perfect stem. It is of considerable interest, however, that in certain lexical items an early channel for the introduction of a finite perfect middle paradigm appears to have been the perfect middle participle. The clearest such case in Indo-Iranian is that of the root jur¿- 'enjoy', which furnishes twelve Rigvedic occurrences of the participle jnjur¿iirtá-, and no more than three of 3 sg.j7ljur¿é, the only other attested perfect middle form; in the active, on the other hand, 3 sg.jnjór¿a, 3 pI.jnjur¿úlp and 3 sg. subjunctivejújor¿ati, etc. appear more than twenty-five times, while the participlejujur¿vlÍn is found only twice. A similar, ifless dramatic, pattern can also be observed in forms such as tiitl'r¿iirtá- (4 x ), tatl'r¿iirtá- (1 x) vS. 3 pI. tiitrr¿úlp (1 x) « trr¿- 'thirst') and bubudhiiná- (3 x ) vs. búbodhati, etc. (3 x) « budh- 'awake'); further· examples are given by Renou, op. cit., ch. 6. Distributions like these are doubtless related to the fact that, descriptively, jujur¿ii~lá-, tiill'r¿iirtá- and bubudhiiná- are merely reduplicated variants of the shorter participles jur¿iirtá- (34 x ), tl'r¿iirtá- (4 x) and budhiiná- (3 x), with which they agree completely in function 49

). The latter forms, though historically associated

49) On the value ofthe aorist participle in -ana- see the references cited by Renou, op. cit., 137.

Page 41: Jasanoff.1978.Stative and Middle in Indo-European

with the root aorist, were evidently reinterpreted in Vedic as formally aberrant members of the perfect system, and were secondarily provided with reduplication in consequence of their stative meaning. For a similar development, compare the anomalous participle sasayaná-, found twice beside regular sáyana-(18 x) (: preso sáye 'lies').

Although the matter will not be pursued in detail here, it may be conjectured that the relative prominence of the perfect middle parti?iple and pluperfect middle in Greek (cf. the pattern ~mxe:: ~¡XTO,

~fLfLoPe: : dfLIXPfL~vo<; : e:'LfLIXPTo, and the discussion by Chantraine, Hist. d1t parf. gr., 54ff.) ultimately reflects a similar introduction of reduplication into the participle, and later the indicative, of the middle root aoristo Thus, it is ' possible that a form such as ~f:l¡fLe:vo<; was originally stative (perhaps meaning 'lying' ; cf. §§ 33, 95), and that in this value it was renewed by its reduplicated counterpart ~e:~f:l)fL~vO<;50). Subsequently, with the systematic extension of the present: aorist opposition throughout the non-finite paradigm of the Greek verb, an opposition could naturally have arisen between ~e:~A'l)fLÉvo<; (stative) and ~A'l¡fLe:vo<; (punctual); the contrast between these forms would then have served as the starting point for the creation of a stative preterite (pluperfect) ~~~A'I)TO beside the aorist ~A'~TO.

§ 72. The purpose ofthis excursus has been to show that in Vedic Sanskrit and perhaps also in Greek, new perfect middles were created in two important ways: by supplying an older perfect active with middle forms, and by providing an older middle root aorist with reduplication. In principIe, either 01' both processes could have played a role in the extension of the perfect middle in early pre-Germanic: it should be noted in particular, however, that the loss ofreduplication in the Germanic perfect system would have rendered the synchronic relationship of the perfect middle and the root aorist especially close, since it had the effect of merging the two categories in all their modal and participial forms 51

).

At least ,three further class III verbs are referrable to roots for which the assumption of a perfect 01' middle root aorist in Indo-European is probable 01' certain; their presents can thus reasonably be derived from perfect middles created in early pre-Germanic, if not lE times:

50) The accentual difference between ~)\~(le:vo<; and ~e:~).:I)(ll;vo<;, of course, must be seconda1'Y in any case.

51) Indeed, the rarity ofmodal forms ofthe perfect middle in Indo-Iranian even suggests the possibility that the subjunctive and optative of the perfect middle were supplied by unreduplicated, i. e., root ao1'ist forms in Indo-European. The early convergence of the two categories may also be responsible for the twofold semantic va.Iue (stative and non-stative) of class III p1'esents like *witan, *-aip, the sense of which in Gothic ('observe') is perhaps best explained on the basis of a middle 1'oot aorist 3 sg. y,idó 'noticed'. Cf. § 101.

82

r ¡

*liban, *-aip 'live' (Go. liban, 01 lija, ORG leb'f'n, OS, OE *libjan): a middle root aorist is directly indicated by the Toch. B class In present lipetii'i' 'remains over', refiecting a 3sg. *lipó(r) (cf. §34).

*wunan, *-w) 'be content, stay' (Go. *wwwn (in 7lnwunands 'dissatis­fied'), 01 una, ORGwonen, OE wunia.n (lI, with traces oflII; cf. Flasdieck, 58)): the Vedic impv. 2 sg. vár¡~sva, attested five times in the Rigveda, should probably be regarded as a root aorist; while a sigmatic interpreta­tion is also possible, the only other Rigvedic instances of an 8-aol'ist imperative in -sva are the doubtful hapax sák<$va, 3.37.7, andrti8va, taken by Narten, 8igm. Aor., 219~20, as a "medialization" of 2 sg. acto 1'tiS1;r,2). Indirect evidence for a 1'001, aorist is also furnished by the class VI opto 1 pI. vanéma ,and' subj. 3sg. vantiti (cf. §§66, 100); note further vantct, RV 1. 139. 10, rendered by Geldner as 3 sg. 'el' beansprucht' rather than as a syncopated 3 pI. (for *vanata 01' *vanantL~)53). The corresponding perfect is found in the Ga,thic Avestan participle vaunus (nt. sg.) 'flehentlich', y. 28. 8.

*pulan, *-a'ip 'endure' (Go. pulan, 01 pola, ORG do len , OE J)olian (lI, with traces of lII: cf. Flasdieck, 56): an lE perfect is indicated by Gk. 1ÚAlXfLe:v, TÉT/\'I)U'¡IX, etc. and OLat. tetul'i; likewise inherited is the 1'001, aorist hAa, although middle forms of the type *(s)T/\aTo ( < *t~hAt)ó) are nowhere directly attested.

A number of other class III verbs may belong here also, although the details of their morphological pl'ehistory are less clear. A representative example is *l71ban, *-aip (cf. Go. lubw:n8 Chope'), the Germanic cognate of Lat.lubere and Ved.lúbhyati 'desil'es eagerly' (AV). No old aorist oftheroot *leubh- survives in Vedic; it is by no means impossible, however, that like other zero-grade presents in -ya-, lúbhyaü was originally associated with a middle 1'001, aorist, to which it stood in the same relation as, e. g., tl'flyati 1,0 tr<$a1]á-, mányate 1,0 ámata and búdhyate to budhaná-.

The o-grade root vocalism ofGmc. *hangan, *-aiJ¡ 'hang (intr.)' (cf. Go. hahan, 01 hanga, ORG hange:n) might at first glance suggest a connection with the pe1'fect for this ver b as wel1. We have already noted (cf. ch. 2, note 42), however, that the related Sanskrit verb, sa1ikate 'doubts, hesitates' was pro bably originally an athematic present; for the prehistoric renewal * sa1ike ~ .~a1ilcate compare the repla,cement of athematic stáve 'is praised' by stávate in historical times. It seems attractive, therefore, to reconstruct a 3 sg. middle *konkor, an indirect trace of which can perhaps be identified in

';2) Narten herself, however, takes vá'J¡¡,sva as sigmatic (p. 235). 53) The root van- has both 8e~ and anit forms (cf. N a1'ten, loe. cit.); vCf.1¡{a, if

genuinely a 3 sg. root aorist, could in principIe represent either an old *llénto (aniO 01'

the replacement of an eal'lier *vana<*y,1Jhx-ó (set). Whethe1' 01' not a middle 1'oot aorist is actually attested in Vedic, of course, the significant fact is that such a form probably did exist in the parent language.

83

Page 42: Jasanoff.1978.Stative and Middle in Indo-European

Hittite (§43)54). The prefürm *fConkol' , üf cüurse, immediate~y invit~s cümparisün with3 sg. *tOl'pOl', the assumed sü~rce üfGm~. *parbafp.(§ 70); lf üur interpretatiün üf this fürm as an ac!ÜstatlC pres~nt IS cOl:r~ct, It wüuld füllüw that the active cüunterpart üf *konkol' was *k6nke, üngma!ly a 3 sg. in *-e üfthe type discussed in nüte 47 abüve. Pro. bable reflexes üf *k6nke can be seen in Hitt. kanki 'hangs (tr.)' and Gmc. *hanhip, Gü. hahip 'id.': nüte that under this analysis the cüntrast between *hanhip and *hangaip wüuld simply reflect an üriginal difference üf vüice 55).

§ 73. A final grüup üf Germanic fürms in which the 3 sg. ending *-aip is müst simply analyzed as *-ai + *-p are the nasal presents with class III inflectiün, the clearest examples üf which are *kunnan, '*-aip 'recügnize' (Gü. ana-, at-, get-, ~tf-k1mnan, ORG kunnen), *liznan, *-a.ip 'learn' (ORG limen, lemen, O E leomian (H, with traces üfIII; cf. Flasdieck, 48-9)) and *Mina.n, *-nip 'lean' (ORG Minen, OE Minian (HjIII; cf. Flasdieck, 47~-8)). AIsü üfthis type, thüugh with unambiguüus third c1ass fürms ünly in Old High German, are *ginnn, *-aip 'yawn' (ORG ginen; cf. OE 3ininn, OS ginon (II)) and *1n~tmnn, *-niP 'grieve' (ORG momen; cf. Gü. mn1.tmnn (stem *ma~tmni-Ima ttested), 01 momn, OS momon (H)); further examples are given by Flasdieck, 140ff.

The süurce üf the class HI paradigm in these verbs, which is c1early üf Cümmün Germanic date (cf. Wissmann, Nominn postverbnlin, 141ff.) is prüblema,tic. J. Schmidt's recünstructiün üf an lE nasal present type in *-na'i- (Festgntj3 Roth, 179), whence in certain envirünments Gmc. *-nni-, can no. lünger be seriüusly entertained. Nor is there any likelihüüd that Gmc. *-naip can be explained ün the basis üf an lE nasal class in 3 sg. *-ntli < *-néh}ti: nüt ünly is it difficult tü envisage hüw an ending üf this shape cüuld have yielded -naip, but there is no. evidence that Indü­Eurüpean ever püssessed such fürms except in nasal infix presents tü rüüts in *-hj (e. g., in *pI-n-éh¡-ti 'fills' < *plehr)fl6). The püssibility that verbs like

54) Similar1y, Lat. c7mctor, -arimay be based o.n a 1o.stdepo.nent *concoT, *-i 'hang (intr.)'. '. .

55) The accentuatio.n o.f *torpoT and *konkor has dehberate1y been left unspeCl­fied. Altho.ugh we sho.uld have expected these fo.rms to. sho.w ro.o.t accentuatio.n, *jJarbaijJ and *hangaijJ clearly co.ntinue o.xyto.ne prefo.rms. It is pro.bably simplest to. suppo.se that at so.me po.int in the prehisto.ry o.f Germanic )nherited *tórpol' and *féónkor, 0.1' their co.ntinuants, were remade to. *torpór and *konkól' o.n the mo.del o.f o.ther athematic middle fo.rms.

Superficially similar to. *jJarban and *hangan are the o-grade class III verba *sagan and *wakan 'wake'. The vo.calism o.f *sagan, ho.wever, may reflect the influence o.fthe iterative-causative *sagjan; that o.f *wakan canno.t be separated fro.m the o-grade o.f the underlying adjective *wakraz (= Gel'. wacker).

56) To. be sure, it is no.t unthinkable that the ro.o.t o.f*kwnnaijJ was o.fthis shape; Lindeman, NTS24, 7-12 (1971), has attempted to. establish *h1 as the fmal laryngeal o.f "classical" lE *gnii- o.n the strength o.f Hitt. gane8zi 'reco.gnizes'. But there is no. po.ssibilíty o.f ro.o.t-fmal *h1 in the o.ther presents in *-naijJ.

84

*lcunnnn üriginally fürmed presents in *-no- ( < *-na- < *-nehz-) which were secündarily transferred tü the third class is remüte; the ünly nasal presents which seem tü have been prüductive in Germanic times belüng tü the inchüative type seen, e. g., in Gü. ga-wa.knan, -nip, pret. -nada 'wake up' and 01 vakna (II) 'id.'.

A characteristic feature üf nearly all Germanic nasal presents is their ingressive 0.1' durative functiün. Müst are alsü intransitive: in additiün tü the fürms already discussed nüte, e. g., such synchrünically primary verbs as Gü. keinnn, pret. keirwdn 'bud, sprüut' beside OE c'innn, ORG ch'ínan (strüng) 'id.', and OE dw'innn (str.) 'disappear' beside 01 dvínn, dvena (II) 'id.'. Früm a histürical püint üf view this situatiün is surprising, since the nasal presents of Indü-Eurüpean seem typically tü have been transitive, and üften, as inHitt.lJ,am'i(n)k- 'destrüy' vs.lJ,nrk(iya)- 'perish', Ved. p7mlÍti 'purifies' vs. pávate 'flüws in a purified state' (of the süma), 0.1' Ved. inóti 'drives' vs. ét1: 'gües', tü have had a specifically causative 0.1' factitive sense. Taken as a grüup, the Germanic nasal forms accürd closely in value ünly with their mürphülügical cüunterparts in Balto-Slavic, where both the Baltic type with infixed -n- (e. g., Lith. bur1da" inf. Mlsti 'awake') and the Slavic type in -nejo- (e. g., OCS vbz-bbnetb, inf. vbz-bbnqti 'id.') are normally utilized tü form inchüatives similar in functiün tü the Güthic ga-waknan type.

In my view the semantic behavior üfboth the Germanic and Baltü-Slavic nasal presents can best be accüunted für by assuming that verbs like 'awake' in these languages were üriginally cünjugated in the middle, and that their attested thematic active fürms represent replacements üf earlier fürms cümparable tü Ved. 3 sg. mido grr.¿e, grrp'ité beside acto grrplÍti < gr­'práise'. The specific mürpholügical prücesses which led tü the establish­ment üf the regular active types in Germanic and Baltü-Slavic can be envisaged in a variety üf ways, and need nüt cüncern us here 57). Of immediate relevance to the present discussiün is rather the fact that the assumptiün üf medial nasal presents für pre"Germanic permits a straight­fürward explanatiün üf fürms like *k1mnaip, *liznaip and *hlinaip as renewals üf older 3 sg. forms in *-nai < *-nhx-ói. In the case üf *k1mnan in particular, the inherited 3 sg. *k1.tnnai « *(J1J-n-h3-6i 0.1' *gr;¡.h3-nhz-ói) 58)

57) Specifically, it is unclear whether the thematic nasal presents in *-nelo­represent simple thematizatio.ns o.f o.riginal active presents in *-nehz-j*-nhz-, 0.1' were directly remade fro.m o.lder middle fo.rms. In either case, the incho.ative functio.n o.f nasal presents in Germanic and'Balto.-Slavic wo.uld have o.riginated in the middle, and beco.me general o.nly after the causative and factitive functio.ns o.f the active hOO been taken o.vér by o.ther fo.rmatio.ns, no.tably the type in *-ejelo-,

58) The latter reco.nstructio.n, certainly the less like1y, wo.uld sho.w the zero.-grade o.f *gnehs- fo.llo.wed by a synchro.nic su:ffix *-nehz-j*-nhz-,

85

Page 43: Jasanoff.1978.Stative and Middle in Indo-European

r----------------I

would originally have stood in the same formal and functional relation to the preterito-present *htnn 'lmows' as did *dugai to the old perfect *dcmg; indeed, it was doubtless on the basis of a proportion of the type *d7lgai : *dwug :: *kunncli : X that *1cann was itself created, replacing an earlier form akin to Skt. jajñau and Lat. gnouit. Gmc. *liznaiJ¡ and the preterito-present *lais (= GO.lais 'knows (how) , understands') constitute a similar pair: here, however, synchronic restrictions on the structure of verbal roots in Germanic ruled out the possibility of a renewal *lais -> **lcásn.

§74. The lexical items discussed in §§62-73 supply the bulle of the evidence for supposing that the paradigm of the third weak class is ultimately traceable to a Germanic replacement of *-ai by *-wijJ in the third person singular. Although these words account for only a comparatively small fraction of the total membership of the third class, they make up a much higher percen tage of the class III ver bs with primary ver bal cogna tes outside Germanic.

It would not, however, be necessary or desirable to attempt an explanation of the numerical majority of third class verbs on the basis of the paradigm reconstructed for *dugan, *-aip in § 65. This is especially clear in the case ofthe denominative statives and inchoatives(cf. §50), which almost certainly continue, directly 01' indirectlYi the lE denominative type in *-e- or *-eielo- (cf. §6). It would be easy to account for the classIlI inflection of, e. g., Go. ana-silan 'fall silent' (: Lat. SaeTe) ifit were possible to suppose that lE *-eie- yielded *-ai- in Germanic: we would then assume that the morphological identity of *silan and *d7lgan in the 2,3 sg. and 2 pI. led to the formal merger ofthe two types everywhere. Since we have seen, however, that a derivation of *-ai- from this source is in fact very dubious (§ 52), it is probably safer to regard the inflectional agreement of the class III denominative and deverbative statives as a simple consequence of their close functional paraHelism. The confusion of the two formations would clearly have been favored by the fact that certain third class verbs belonging historically to the *dugan type might easily have created the synchronic appearance ofbeing denominative. The ambiguous position of *pm'ban, *-clip in this regard has already been noted (§ 70); a similar case is that of *s(w)'n1'gan, *-Clip 'worry' (ORG s(w)o1'gen, OE S01'3ia.n (Flasdieck, 49), probably Go. sa7lTgan) , which, although probably based directly on the root *s(JI)er-gh- 'worry, be sick' (Pokorny, IEW, 1051), could have been interpreted within Germanic as a derivative of *S(W)7lr-gO 'sorrow' (Go. Sa7lTga, 01 S01'g, ORG s(w)o1'ga, etc.). A close typological parallel to the replacement of denominative forms of the type *silep ( < *-e(ie)t?:) by new forms in *-aijJ is provided by the occasional cases in Baltic in which a denominative stative shows the characteristic i-inflection of the deverb­ative class (cf. Lith. gáilis 'feels sorry', inf. gail¿Us<gailus, sf¡1csti 'is

86

stingy', inf. syksteli < SyICSt1lS) in place of the regular denominative paradigm in *_~a_59).

More obscure are the factitive denominatiolls of the type seen in Go. swemn, -aip 'honor', ga-ainan, -aip 'make separate', etc. (cf. §50). In a recent study of these forms (Lg.52, 851---65 [1976]), J. Dishington has attempted to show that they continue an lE present class in *-oielo-, which he also takes to be the source ofthe Greek present typein-ów (30uA6w, xEv6w, etc.). The patently secondary character of the latter formation, however (cf. Schwyzer, GT. G1'.727), deprives the alleged lE parent formation of serious extra-Germanic support 60

). It is at least equally possible, in my view, that the class III factitives represent secondary transfers from the second we!1k class, where the formation represented by ORG mihhilan 'make great', meTan 'make more', etc. is of lE date (cf. Ritt. newalj- 'make new'). The twofold value of forms like *fastan, *-aip (cf. Go. fastan 'hold f'irm' and 'fast' , ORG fasten 'fast') and *piwan, *-aip (cf. Go. ctna-piWCln 'subject', OE peowian 'serve' (Flasdieck, 54-6); 01 pjá 'serve') would then reflect the earlier presence oftwo independent verbs, a class II factitive and a class III stative.

Denominatives, it should be noted, do not necessarily exhaust the cases in which a morphological replacement of *-ep « *-e(ie)ti) by *-aip can usefully be assumed. The only clear extra-Germanic cognat)e ofGo. pahan, -aip 'be silent' and ORG dagen 'id.' is Lat. tace Te (cf. Umbr. taQez 'tacitus'), to aH appearances a primary e-verb; since there is no evidence that the root *tak- formed a perfect 01' root aorist in the parent language, it may be simpler to reconstruct a late lE *takitti 01' *takitie(ti) than to postulate an actual 3 sg. middle *takói as the ultimate source of Gmc. *pagaip 61). Similarly, it is difficult to find support outside Germanic for deriving *habaip 'has' from a 3 sg. middle *1capói; an original *1cap~ti 01' *1cap~ie(ti), on the other hand, would accord perfectly in structure and formation with Lat. habet 62

). The lE distribution of *-e- in non-denominative statives is discussed in § 105.

59) It is true that the denominative and deverbative statives of Balto-Slavic agree in having stems in *-8- outside the presento But it is at least possible that a similar situation once existed in Germanic, and in any case, the semantic similarity of the two types would alone have been sufficient to bring about their formal merger.

60) The Greek type in -6(,) appears to have been back-formed from the adjectival type in -(,)T6" the basic pattern having been supplied by presents in -6:(,) with verbal adjectives in -aT6,.

61) The reason for the elimination of gmmmat-ischel' Wechsel in Go. pahaip is not clear to me,

62) Lat. habel'e, of course, presupposes a root-form *ghabh- rather than *7cap-. A genuine example of a perfect middle meaning 'have' is Ved. tse, akin to the Germanic preterito-present *aih, pI. *aigum.

87

Page 44: Jasanoff.1978.Stative and Middle in Indo-European

§ 75. Before summarizing the results of the preceding discussion, i~ will be useful to return briefly to the formal analysis of the class III paradlgm. In this and the following sections we shall see that the explanatiOI~ proposed in § 65 for the present indicative of the third weak class leads to a straightforward explanation of the non-indicative and non-present forms as well.

No special problems are pl'esented by the third class modal forms. The inherited optative type is preserved in Go. habcw" -ais, -cti, etc., 01 haja, hajir, haji and ORG habe, habes, habe; the latter forms, it will b~ note~, constitute additional evidence for assuming that in West Germamc, as m Gothic and Scandinavian, the class III suffix originally had the shape *-ai-j*-a- rather than *_ai_j*_ja_ 63

). This inflection is ~impl:y that of a normal thematic optative, and is doubtless to be explamed m the same fashion as forms like 1 sg. *habo (*dugO'¡ 01' 3 pI. *haband (*dugand): outside the 3 sg. and its immediately associated forms, the athematic middle endings which characterized the ancestor of the third class paradigm were mechanically replaced by their thematic active counterparts. In t~e imperative, the 2 pI. is the same as the indicative, as everywhere else m Germanic, while the 2 sg. impv. (cf. Go. habai, 01 vctki, ORG habe) is a new form, evidently created according to a proportion of the type 3 sg. *hanz'ip 'hears': 2 sg. impv. *ha1tz'i:: 3 sg. *salbop: 2 sg. impv. *salbo:: 3 sg. *hctbctijJ: 2 sg. impv. X, X = *habai 64

).

In principIe, the class III present participle in *-and- can be explained in the same ,vayas the optative: there is no reason why forms like *dugand-, *witand-, etc. could not have replaced earlier participles of the type found, e. g., in Ved. dúhiina- (d1thiiná-), vídiina- (vidiiná-), etc. In fact , however, there is a second, and simpler possibility. In ternal evidence from Vedic suggests that in certain cases middle participles in -iina- haye either replaced 01' encroached upon the domain of earlier verbal adjectives in -nt-, traces ofwhich can be seen in such pairs as dhrf!ánt- : dh'(f!ii'f!á- (AV) < dhrf!­'dare', vrdhánt- : vrdhiiná- (viivrdhiiná-) < vrqh- 'increase' and, especially, duhánt- (cf. § 64): dúhiina-, duhiiná- < dnh- 60). Sínce the assumption of a pre-Germanic stem *dh11.ghént-j*dh'ughr,d- is necessary in any case to account for the feminine abstract OE dU3uP, ORG tugund 'yirtue' < *dhugh'{ltá, it is attractiye to consider the possibility that the class III participle *d1lganel- is directly cognate with Ved. eluhánt-, from which it differs significantly only in the yocalism of its second syllable66

).

63) Ifnecessary we could as sume an analogical replacement *habjai- --+ *habai- to account for the OHG optative, but a direct explanation is cleady preferable.

64) In Gothic the *-cd of this form has been retained under the morphological influence of the rest of the paradigm.

65) It is doubtful that any special significa.nce should be attached to the fact that duhánt- is transitive; so too, on occasion, is d7lhaná- (dúhana-).

66) Tha t is, I assume tha t this and similar ver bal adj ectives wi th zero-grade of the root had a declension of the "hysterokinetic" type.

88

~---------------------------------------------------------

The status of "active" nt-participles beside middle finite forms has been discussed in § 39.

§ 76. Eyidence corroborating this interpretation is proyided by the unique R unic N orse compound witaclahalaiban (Tune, c. 500 A.D.). Accord­ing to the prevailing and doubtless correct view, this form represents the dato sg. of an n-stem *witand-hlaiban-, itself built, via a productiye Germanic process, to an underlying *witand-hlaiba- 'watching oyer the bread'. The sense of the n-stem is thus 'one who watches oyer the bread, lord' , showing the same semantic deyelopment as O E hliijorel 'lord' < *hliif­weard 'bread-guard' 67). Compounds ofthis kind, in which the first term is a participl\l whi(lh goyerns the second, are otherwise unlmown in Germanic; they are, howevel', well-attested in the oldest Indo-Il'anian, where fol'ms of the type Ved. bharádviija- beal'ing the pl'ize' (cf. Ay. bamt. zao8m- 'beal'ing the libation'), elham,yátkavi- 'protecting the wise', etc., constitute a sizeable class. Especially conspicuous among the pal'ticiples found in such com­pounds is IIr. *vidát-, which occul's in vidádva,s'u- 'finding wealth', *vidádasva- (in váidadasvi- descendant ofV.') 'finding horses' (PN), YAyo v'idat. gauu- 'del' del' Kuh teilhaftig wird, del' für das Rind sorgt' (Bartholo­mae) and vioat. x"arCJncth- 'del' des XV. teilhaftig wird' (PN). The complete isolation of *witctnd-hlaiba- in Germanic makes it probable that both this form and the Indo-lranian vidát-compounds continue a single lE type with *y,iel1;,d-in initial position 68). Sínce, howeyel', Gmc. *witand- in composition is formally and semantically indistinguishable f1'om *witand-, the participle of *witan, *-aip (Go.: 'observe, watch'), the conclusion natul'ally suggests itself that lE *ltidént-j*y,idr,d- is the source of thelatte1' fol'm as we1l 69

).

These obseryation afford an insight into the actual origin of participle­object compounds in Indo-European. In outward appearance 8uch com­pounas resemble normal bahuvrihis such as Ved. dhrf!ádva1'1.w- 'having a bold nature', '/'1tsátpa!nt- 'haying bright cattle', s1LCádmtha- 'haying a shining chariot', brháel1Lkf!an- 'having tall bulls', etc., in which the first element is an nt-stem. The lE type *lLidr;.t-X, in my yiew, originally had precisely the same structure as cases ofthis latter kind, and properly meant 'having an X which is found (seen, lmown), . The possibility that *llidént-j *llidr;.t- had passive as well as active functions in Indo-European, of course, is suggested by the passive yalue of nt-participles in Rittite; in Vedic, v'ídiina- (vidiiná-) , the replacement of *vidánt- as a free form, may mean

67) It is hardly possible that the participle of a class II *witon could have entered into a compound ofthis type, or that it would have been spelled witada- in A. D. 500.

(8) As Schindler has pointed out in a personal communication, the zero-grade *llid~lt- in Iranian as well as Indic ex eludes the possibility that *vidát-belongs to the paradigm of the thematic aoristo

69) It is less attractive to suppose that the participle *witand-represents a wholly new creation, to which *wihmd- ( < *ll'Íd1Jt-) was secondarily assimilated at a later periodo

89

Page 45: Jasanoff.1978.Stative and Middle in Indo-European

~~ __________________________________________________ au7.nr"up2·"5"""""""""" .............................. 111111 .... 111111. _ .... __ .~~ ---'known' as well as 'knowing' (> 'wise'). The syIltactic reinterpretation of *uidnt-X as 'finding (seeing, knowing) x' may already have begun in late I;;'do~ European; in Inclo-Iranian *vidát- was synchronically identified with the 3 sg. thematic injunctive *vidát ancI new compounds of the same type were created from other roots with thematic presents 01' aorists70

).

§ 77. The preterite system ofthe third class, like the present, is markecl by formal cliscrepancies among the attestecllanguages. The "regular" preteri­te type is founcl in GJthic, Scanclinavian ancl Olcl High German, where the clental suffix of the weak preterite is separatecl from the root by a connecting *-ai-; this vowel is preserved in Gothic (habaida, habaijJs) ancl Olcl High German (habe:ta, gihabe:t) but is usually syncopated in Olcl Icelanclic(hafoa, hafor, nt. (arch.) hafat). Old Saxon ancl Old English, on the other hancl, attest a second type in which no union vowel is present (OS habda, gihabd, OE hoefda, (3e)hoefd); a trace ofthe same formation can also be seen in OHG hapta. Flasclieck's opinion (p. 160) notwithstanding, there can be little cloubt that the latter pattern is the more archaic, for while it is difficult to see how a form such as 1 sg. *habaido" coulcl have been replaced by *habdon in West Germanic, the pattern *nazijJ: *nazido" « *nazjan 'save'), *salbojJ : *salbOdon woulcl have proviclecl an obvious :n:0clel for the creation of *habaido" from 3 sg. *habaijJ. Inclependent eviclence for the relative antiquity of *habdon ancl similar forms is provided by the Gothic third class verb uf-kunnan 'recognize', the preterite ofwhich is repeatedly attested as uf-k unjJa , but appears once in the regularizecl shape uf-kunnaida (1 COl'. 1. 21). The existence of a formerly productive process whereby weak preterites ofthe form * X-dO" were replacecl by new forms of the type *X-aidO"' may also permit an explanation for the ex­ceptional preterite of *hugjan 'think' : here an original *hugdO"' (OS hogda, OE h03de, OHG hocta) was evidently renewed by *hugaidon (OHG hagaa, OGutn. hugjJi, 01 ptcp. nt. hugat) despite the absence of classln present forms in any attested Germanic language71

). .

'rhe pro blema tic voiced clusters of * habdon and * hugdon (eL also OS libda, sagda < *libdon

, *sagdo") are probably of analogical origin: under the

70) The lE inventory ofpassive nt-participles which figured in such compounds was probably not confined to *Ltidént-j-1}i-, though the proliferation of the type in Indo-lranian seems clearly to have followed the reinterpretation ofthe first term as transitive and active. vVhether the syntactically similar formations represented by Gk. &.pXÉ){COW~ (cf. Ved. tmsádasYlt- 'affrighting foes') and Tep~((J.~pOTO~ (cf. Ved. dilUvara- 'granting wishes') originally had a passive first element remains to be determined.

71) Go. pret. bc¿uaida, if not simply the regular preterite corresponding to a lost *bcmaiP 'dwells' (the attested form is bauiP) may have arisen in the same way; the prehistory of this verb is obscure. Note also OIflóa 'flow' and óask 'fear', with class III forms in the preterite only.

90

I

influence of the root-final *-b- and *-g- which characterized the presents of these verbs the phonologicallyregular groups *-ft- and *-ht- were apparent­ly remade to *-bd- and *_gd_72

). This peculiarity aside, it seems legitimate to conclucle that the weak preterite of the third weak class was originally made by suffixing the appropriate dental endings directly to the verbal root-a finding which accords perfectly with our view tha,t *-ai-, in historimil terms, represents not a suffix but a desinence.

§78. The origin of the Germanic dental preterite is probably also to be sought in an athematic middle forlllation. Although the full range of evidence favoring this conclusion will not be presented here, the history of the weak preterite is potentially significant fOl' a correct understanding ofthe formal history ofthe third class, and it is in this connection that the following theory is briefly offered.

Both preterito-presents and their associated class nI verbs, in my view, originally utilized middle root aorists of the type 3, sg. *mrptó, ~r reduplicated (pluperfect) forllls of the type 3 sg. *memrpto, to lllake thmr preterites. Thus, after the loss ofreduplication in pre-Germal1ic, a represen­tative group of presen t al1d preterite forllls of the verb tha t was even tually to becollle *munan, *-aip (nI) would have been the following (cf. ~ 65) n): preso sg. 1 *munai pret. sg. 1 *muna (cf. Ved. -1:)

2 *mundai 2 *m~mda ( cf. Ved. -th?il¡,) 3 *munai 3 *m~mda (cf. Vecl. -tCt)

pI. 3 *nmn7mdai pI. 3 *mummda (cf. Ved. -ataf4) opto *m7m'i- opto *m1m'i- (cf. Ved. -i-)

It may 110W be conjectured that, as occasionally elsewhere in Indo­European75

), the oppositiol1 betweel1 the primary al1d secol1dary llliddle

72) 01 olla>wltlpon, 1 sg. pret. of va.lda 'wield', provides one of the strongest pieces of evidence foI' supposing that the dental of the weak preterite was *t. The often expressed view that the weak preterite contains a form of the verb 'do' ~s refuted, as we shall see below (cf. especially note 76), by the fact that the Rumc Norse 3 sg. in -de (j({hide, tawide) continues not CGmc. *-de but *-da/i.

1 am indebted to Cowgill for having pointed oüt to me that the absence of a ~ental in ONorw. pret. sg. haje and similar forms (cf. Noreen, Altnord. Gr. 1, 350) lS the result of a, secondary phonological development. The Common Germanic prototype of haje must be reconstructed *habdai. .

73) My decision to assume that the basic sound change~ of Germamc ~lac~ al~'eady operated at this stage is, of course, arbitrary. The ana~ys1s offered .he1'e lS slmIl~r to that set forth by H. Collitz in Das schwache Pmetent1tm 11nd seme Vorgesch'Whte (1912); a different, but closely related interpretation will be presented in a forthcoming publication by Hollifield.

74) 01' conceivably *l1mnm, although the presence óf a dental ending in the 3 sg. does not favor this possibility.

75) Compare, for example, Osc. 3 sg. -ter, 3 pI. -nter in primary an.d secondary functions, beside the more archaic Umbrian opposition -ter; -nter (prIm.) vS. -tll,r, -ntu,1' (sec.).

91

Page 46: Jasanoff.1978.Stative and Middle in Indo-European

endings was neutralized in favor ofthe former set in Germanic, leading to the (originally optional) introduction of final-ai into the paradigm of the preterite. Outside the 3 sg. this change would have had the effect of introducing an undesirable homophony between the two tenses; the resulting situation could have been rectified, however, by generalizing the distinctive difference between the 3 sg. preso *munai and 3 sg. pret. *mundai, i. e., the absence vs. presence of a dental, to the rest of the paradigm. The preterite forms accordingly acquired a dental everywhere, while the dental ofthe 2 sg. was dropped from the presento The new system would have been:

preso sg. 1 *mnnai pret. sg. 1 *ni1tndai 2 *munai 2 *mnndai 3 *mnnai 3 *mundai

pI. 3 *mnmtndai pI. 3 *rmtndundai opto *m1tn'i- opto *m1tnd'ir-

The three persons of the singular were now alike in both tenses. With the decay of the middle as a category a natural expedient for distinguishing the 1 sg. preso from the remaining singular forms would have been the substitution of *-0 for the obsolescent middle ending *-ai; given the systematic relationship between the present and preterite paradigms, moreover, such a renewal could easily have led tothe creation of a corresponding 1 sg. pret. * mundo. After the remodeling of the 1 sg. the present and preterite appear to have developed along independent lines. In the present the loss ofmiddle inflection was marked by the replacement of 2, 3 sg. *m1tnai by 2 sg. *m1tnais and 3 sg. *rmtnaip, and by the creation of thematic forms in the 1, 3pl. and optative on the model of 1 sg. *nwnoand ptcp. *nntnand-. In the preterite the middle endings of the plural and optative were replaced by their active (strong preterite) counterparts, while the singular forms of the first and second persons were further extended by the corresponding secondary active endings. The final result would have been a system like the following, which may be assumed for Common Germanic 76) :

preso sg. 1 *1mtnO 2 3

pI. 3 opto

*m1tnais *munaiP *munanp *munai-

pret. sg. 1 2 3

pI. 3 opto

*mundo" *1n1tndais *mundai *mund1tn *mund'i-

76) As a final step, of course, *nw,nda¡; was replaced by *m1lnaidai under the influence of the associated present stem. The ending *-dai in the 3 sg. is assured by the archaic form talgidai 'engraved' (N 0vling, third century); Hollifield will show separately that the conventionally reconstructed *-dewould have yielded Runic *-da. The Gothic weak preterite forms in -des, -de:dum, etc. appear to reflect a secondary analogy with the preterite of 'do'.

92

"

Thus, three steps can perhaps be distinguished in the genesis of the class III pattern: 1) the replacement of 2 sg. *-dai by *-ai, resulting in the establishment of *-ai as the singular ending in all three persons; 2) the replacement of 1 sg. *-ai by *-0; and 3) the replacement of 2,3 sg. *-ai by *-ais, *-aiP, leading to the eventual generalization ofthe suffix *-ai-j*-a­throughout the paradigm.

§ 79. The preceding discussion has presented arguments to show that the third weak cIass is based not on an lE present type in *-e-, *-ejejo-, *-ei- 01'

*-ojejo-, but on a group of inherited athematic middle formations charac­terized by a pre-Gmc. 3 sg. in *_oi 77

). Important for this interpretation was the existence' of an archaic group of third class verbs associated with preterito-presents: in such cases an original middle paradigm was suggested by the derivational affinities of the perfect with the middle in Indo­European. One verb of this group (*dugan, *-aip; §§ 64-5) has been refened to an older zero-grade root present, while two C'thers (*munan, *-aip and *witan, *-aiP; §§ 68;-9) have been traced to earlier perfect middles; such verbs as *wunan, *-aip and *pulan, *-aiP (§§71-2) may continue perfect middles as well. Gmc. *parban, *-a.iP (§70) and *hangan, *-aiP (§72) have been explained as reflexes of an acrosta tic presen t type with al terna ting *0-and *e-grade vocalismo The nasal type represented by *htnnan, *-aiP and *liznan, *-aiP (§ 73)points to an originallydeponent cIass in 3 sg. *-nh"ói; an aorist present like those discussed in ch. 2 appears to underlie *fulgan, *-aiP (§ 64). We have surmised that the inflection of the class III denominatives (*silan, *-aiP, etc., § 74) reflects a process of secondary assimilation to the statives of the deverbative type.

As we shall see below, this analysis can be pursued further: in § 101, it will be suggested that the lE antecedents of Gmc. * dugaiP , *fulgaip and *munaiP originally constituted a single morphological category, rather than three distinct formations. Before speculating on the Indo-European sitnation, however, it will be useful to investigate the position ofthe stative presents of a third branch of the family, Balto-Slavic; here, as the following chapter will show, the Baltic type in -i- and the Slavic type in *-'i- present a historical picture strikingly similar to that of the third weak class in Germanic.

77) My earlier analysis of the third weak class as a reflex of the thematic middle (Lg. 49, 85{}--70 [1973J) while obviously closely related to that presented here, was partly based on an incorrect appraisal of Gmc. *hangaip, which l no longer regard as continuing an lE thematic presento Whether any originally thematic middles have found their way into the third class is doubtful; theirregular treatment in Germanic has been discussed in §46.

93

Page 47: Jasanoff.1978.Stative and Middle in Indo-European

IV

STATIVE PRESENTS IN *-t'- IN BALTO-S LA VIO

§ 80. The functional position of the third weak class iri Germanic is partly occupied in Balto-Slavic by verbs with an infinitive stem in *-e- (Lith. -e-, OCS -é-) and a present stem in Balt. *-i-, SI. *-i- (Lith. -i-, OCS -i-). The forms of *mingti 'think, remember'l) in Lithuanian and Old Church Slavonic may be taken as representative of the class as a whole:

Lith. OCS

preso sg. 1 miniu mbnJq 2 mini mbnisi 3 mini mbnit'b

pI. 1 m1nime mbnim'b 2 minite mbnite 3 (mini) mbn~t'b

imp. sg. 2 miné1c mú¡,i 3 te-minie mbni

pret. sg. 3 mine,jo mbné (aorist)

ptcp. pret. acto minej~s, minej1ls- mbnév'b pass. minéla,s mbnén'b

parto preso acto minjs, minint- mbn~, mbn~st-pass. minimas mbnim'b

inf. minéU mbnéti

As regularly in Old Church Slavonic, é is replaced by a after a soft consonant; cf. slysati 'hear' for *slySéti, lezati 'lie' for *lezéti, etc.

1) For convenience the ending of the Balto-Slavic infmitive is here reconstructed *-ti, although the OCS infmitive in -ti presumably continues an immediate prototype in *-tei. EIsewhere in this chapter 1 have taken similar liberties with certain of the Balto-Slavic personal endings.

94

.~ i

§ 81. Most verbs of this type (Senn's conjugation class II in Lithuanian, Leskien's class IV. 6 in Slavic) are primary, i. e., deverbative statives; as we have seen in §§ 6-7, it is characteristic of Balto-Slavic that e-verbs of denominative origin are associated with an inflectional pattern oftheir own (cf. Lith. senéU, 1 sg. senéju 'grow older', OCS sta1'éti, 1 sg. sta1'ejq 'id.'). In neither branch are lexical items like *mingti productive as such, although the Lithuan~an iteratives in -séli, 1 sg. -siu (e. g., éepséli 'smack the lips', blikséti 'glitter', str-akséti 'hop about'), intensives in -deti (e. g., mé1'deti 'lie dying', své1'deti 'stagger') and "diminutive perfectives" in pa- ... -éli (e. g., pabégéU 'run. a little and stop', palüké1i 'wait a little') belong descriptívely to the same conjugatíonal class. The inflectional pattern of *minüi is found throughout Slavíc, but in Baltíc ít is well-represented only in Lithuanían. Standard Latvían has replaced the historical *-i- of the present stem by (palatalizing) -a-, normally the reflex of CBalt. *-ia-; some Latvian dialects, however, retain -im and -ü in the 1 pI. and 2 pI., respectively. A similar loss of i-inflection is found in many dialects ofLithuanian 2). In Old Prussian the only clear representatives of the Lithuanian minéti-type are the verbs tU1''it, preso 3 tun'i 'have to, be obligated' (: Lith. tU1'éli 'have') and lcinlit, preso 1 sg. k'i1'dimai 'hear' (cf. Lith. gi1'deJi 'id.'), other eji-verbs having been assimilated to the pattern ofthe e-denominatives (cf. bude < *bude¿ja beside Lith. bUdi 'is awake', mile < *mile¿ja beside Lith. myli 'loves').

The Baltic and Slavic forms in question' agree so nearly that a close historical connection between the two types may be assumed at the outset. Yet, as noted in § 7, the nature of this relationship is quite unclear. Balt. *~e- and SI. *-é- can be traced easily enough to BS *-e-; Balt. *-i- and SI. *-i-, however, can hardly be directly equated, since the former is short while the latter is etymologically long. Over the past century much scholarly effort has been expended in attempts to resolve this difficulty, along with the more general problem presented by the synchronically athematic appear­ance of the Baltic and Slavic paradigms. In summarizing the principa,J theories that have been put forth to accoullt for the ejt'-class, it will be useful to distillguish three main groups of approaches: 1) those which attempt to relate BS *-t'- to the lE suffix traditionally reconstructecl as *-iejo- (the "semithematic theory"); 2) those which trace BS *-t'- to an earlier dissyllabic morpheme *-eiejo- 01' *-iiejo- (the "contraction theory"); and 3) those which derive Balt. *-i- and SI. *-i- from apophonic variants of an athematic suffix, which may be variously reconstructed (the "ablaut theory") 3).

2) See the survey of dialect forms in Zinkeviiíius, Liet. Dial., 341-3. 3) This classification, of course, is ollly adopted for reasons of convenience. In

sorne cases the assignment of an approach to one 01' another group is arbitrary; Schmid's theory, for example, relates the attested forms in -f- both to the traditionally reconstructed class in *-je/o- and to a hypothetical athematic type in *-ei-/*-'Ji-.

95

Page 48: Jasanoff.1978.Stative and Middle in Indo-European

§ 82. An outline of ~he .selnithem~tic theory has alread~ be~n presente.d in §9. According to thlS V16W, promment adherents ofwhlCh mclude MeIllet (see, e. g., his Slnv. com., 232~) and Stang (Sl. ~t. bnlt. Vb., 23ff., Vergl. Grnm. d. bnlt. Spr., 320-2)4), the intransitive and stative jejo-presents of lndo-lranian, Greek and other lE languages were originally only partly thematic, exhibiting an alternation between a suffix-form *-jo- in the 1 sg., 1 du., 1 pI. and 3 pI. and *-i- 01' *-i- in the remaining indicative forms. Semithematic inflection was then allegedly replaced by the fully thematic type.in *-jejo- in the lE languages ofthe southeast (notably lndo-lranian and Greek), while *-jejo- was absorbed by *-i-j*-jo- and *-i-j*-jo- in a dialect area including Italic, Celtic and Germanic. Only in Balto-Slavic did the semithematic conjugation survive in its original stative~intransitive func­tion, the subtype in *-i- being generalized in Baltic and that in *-i­prevailing in Slavic.

The shortcomings ofthis analysis have been pointed out in ch. 1. Outside Balto-Slavic the assumption of a semithematic conjugation is entirely redundant, since, as noted in § 9, the Italic, Celtic and Germanic present types in *-i-j*-jo- and *-i-j*-jo- can as easily be traced to thematic prototypes in *-jejo- and *-ijejo-, respectively. It is strilring thatno sign ofa semithematic type survives in Hittite (cf. pnrkiyn- 'go up', lsnrkiyar 'perish' and similar forms, all fully thematic) 01' Tocharian, where dialectal idiosyncrasies ofItalic and Celtic often reappear. Even ifwe admitted the possibility that Balto-Slavic inherited semithematic presents in *-It-j*-jo­from lndo-European 5), it would be difficult to account for the attested distribution of thematic and athematic forms, since in point of fact neither Baltic nor Slavic provides any evidence at all for the suffix-alternant *-jo­except in the thematic 1 sg. *minjo (Lith. miniu, OCS nthn"jq). Note especially that the preso 3 pI. (OCS mhn~t'h) and present participle (Lith. minint-, OCS mhnr;st- < *mininti(n)-) show *-t'- for expected *-io-: to account for these forms under the semithematic theory it would be necessary to assume that contrary to the prevailing Balto-Slavic pattern (cí. OCS 1 sg. vedq '1 lead', 3 pI. vedqt'h, ptcp. vedqst-, vs. vede- elsewhere in the present), inherited *minjnnti and *minjnnt- were specifically distanced from the 1 sg. and assimilated to the athematic forms in *-t'- 6). Similarly, in the inherited optative (= imperative), OCS 2,3 sg. mhni, 2pI. mhnite, etc. would have to be explained as replacements of earlier thematic *minjnis, *-nit, *-nite, the regular OCS reflexes of which would have been

Here and below, the notation "BS *-f-" should be understood only as a cover symbol for the prototype(s) of Balt. *-i- and pre-Sl. *-i"-. It should not be taken to imply that Balto-Slavic had a single suffIx *-t- with long and short aIlomorphs.

4) Following Brugmann; cf. § 9. 5) It will be recaIled that *-je- yielded *-ja- in Baltic; cf. § 42. 6) l use *a to represent the Balto-Slavic prototype of Balt. *a, SI. *0.

96

'1 I

*mhn'ji, *mb'iíjite. For the corresponding optative forms in Baltic, which present comparable difficulties, see § 98.

In Lg. 32, 520~, P. Tedesco has presented strong arguments to show that OCS u-mb'i'q '1 will die', 3 sg. u-mhret'h should be read with a palatalized r, i, e., as *u-mh'fjq, *u-mhr"jet'h. If this interpretation is correct, forms like u-mhret'h would constitute decisive inner-Slavic evidence for supposing that the traditionally reconstructed intransitive type in *-jejo- (cf. Ved. mriyáte, Lat. morior) was in fact fully thematic from lE times.

The semithematic theory, in short, not only presupposes the existence of a morphological class found nowhere else in lndo-European, but also requires the assumption of a series of analogical changes quite inconsistent with the general structure of the Balto-Slavic verbo Despite the authority ofthe distinguished scholars who have espoused it, its relative popularity is quite undeserved,

§ 83. A very different analysis, although its point of departure is likewise a familiar thematic present formation, is that of Specht, KZ 62, 29-115 pnssim (1935), who derives both Balt, mini- and SI. mhni- from a single lE type in *-eiejo-, which he also takes to underlie the -e- of Lat. 2 sg. taces, rubes, etc.7

). From a phonological point ofview this theory is superficially attractive: lE *-eje~ can plausibly be identified as the source of -i- in the Slavic iterative type in -iti (cf. OCS nositi 'carry', preso nosq, nosisi, -it'h; prositi 'ask', preso prosq, prosisi, -it'h, etc.), and unaccented *-i-, when resulting from a contraction, appears to be shortened in final syllables to -i­in Lithuanian (cf, brólis 'brother' < *brálijns vS. gnidfJs 'rooster' < *gnidíjns). Nevertheless, the objections to assuming an lE deverbative stative type in *-ejejo- are very serious. Specht's derivation ofLat. taceo, -es from *tnkejejo- depends crucially on his derivation of seneo, -es from an earlier *senejejo-; as we have seen in § 6, however, seneois almost certainly to be referred to an earlier * sene(jejo)- rather than to a denominative ofthe type seen in Gk. qnMw 01' Ved, nmitmyáti 'ishostile's). Apart from the Balto­Slavic forms in question, no lE language unambiguously attests *-ejejo- in a deverbative stative function; even within Balto-Slavic itself, Specht is obliged to make the arbitrary assumption that *-ejejo-, which originally characterized both the primary and denominative stative types, was subsequently replaced in the latter forms by *-ejejo- under the analogical influence of the *-e- of the infinitive stem.

To be sure, it is possible to envisage an alternative form of the contraction theoryunder which Balt. *-i- and SI. *-i- would be traced not to lE *-ejejo-, but to a suffix-form *-ijejo-. Lith. m~ni and OCS mhnit'h could then be referred to the same preform as Gk. (J-CdVET()(¡ and Ved. mányctte: in

7) So earlier Bezzenberger, BB26, 171ff. (1901). S) Ved. sanayá- 'ewig sein', however, does not eonstitute evidenee for an lE

* sene(jelo)-, sine e it is pro bably a direet derivative of sána 'von je her'.

97

Page 49: Jasanoff.1978.Stative and Middle in Indo-European

Balto-Slavic we would as sume that the long Sievers variant *-ijejo- was generalized as a morphological replacement of earlier *-jejo-. Such an explanation would be free of a major defect in Specht' s theory; on the other hand, a fal' more natural development would have been the extension of *-ijejo- in the non-stative jejo-class, where roots with a heavy sylIabic structure are verynumerous (cf. Lith.júosti 'gird', 3 p.júosia, veflcti 'weep', 3 p. veflcia, OCS pisati 'write', 3 sg. piset'b, lizati 'lick', 3 sg. lizet'b), rather than in the stative type, where the great majority ofverbs have short 01'

only secondarily long initial syIlables (cf. Lith. htrl3%i 'have', guléti 'lie', OCS dr'bzati 'hold'; with morphologicallengthened grade (cf. § 95) Lith. tyleti 'be silent', lüzeti 'be broken', OCS visé ti 'hang'). .

Both this analysis and Specht's, moreover, are suspect on othe!' grounds. Neither *-ejo- nor *-ijo- can be supposed to have yielded *-i:- in Balto­Slavic; consequently, the 3 pI. in *-inti, present participle in *-int- and optative in *-i- present the same difficulties under the contraction theory as under the semithematic approach (cf. aboye). Even more serious is the fact that the shortening of *-i:- to *-i- observable in Lith. b1'ólis vs. gaidys is unlikely to be of Common Baltic date. The phonological history of the Baltic nominal type in *-ija- has been discussed in detail by Stang, Vergl. Gram. d. balto Sp1'., 188ff. (with literature). The development ofunaccented *-ys to -is in the nominative singular is found ayer only a part of the Lithuanian dialect area, and is completely absent in Latvian, where bréllis 'brother' continues an immediate preform *bmli:s (Balt. *-is yields Latv. -s; cf. avs 'sheep' = Lith. av1,s). While it is logically thinkable that the Standard Lithuanian contrast between '-is and -ys reflects the original Baltic situation, and that Latvian and those Lithuanian dialects which exhibit forms of the type brólys have eliminated an earlier opposition, this possibility is in fact very remote. If the contrast between *'-is and *-[s were ofCommon Baltic antiquity, it would be difficult to explain why leveling, where it has occurred, has invariably been in favor ofthe accented form of the ending, despite the fact that the paradigm of the ija-stems otherwise shows several secondary poin ts of agreemen t with the i-stems (cf. Lith. acc. sg. -1, loco sg. -yje, nomo pI. -ys), where the nominative singular historically ended in *-is. In view of the inability of -is to maintain itself under favorable circumstances in the paradigm of ija-stems, it may seriously be doubted whether a Common Baltic· shortening of 3 p. *mini: to *mlni, even if phonologically regular, would have been sufficient to bring about the complete replacement of forms like 1 pI. *mini:me, 2 pI. *mini:te by mlnime and mlnite 9

).

§ 84. A radicalIy different interpretation of the contrast between Balt. *mini- and SI. *mbni- is offered by the numerous versions of what we have

9) The handful of Latvian forms in -i:- cited by Endzelin, Lett. Gr., §629, have been shown to be secondary by Schmid, Stud. Z. balto U. sl. Vb., 89.

98

-j'.­¡

termed the ablaut theory. Despite many differences of detail, most such explanations agree in supposing that at an early stage ofBaltic and Slavic the ej¡-type was represented by a single paradigm in which strong forms containing an i-diphthong (*-ei- 01' *-ei-) contrasted with weak forms containing the zero-grade alternant of the same suf[¡x. Slavic is then assumed to have generalized the diphthongal forms, in which *-ei- (*-ei-) was regularly monophthongized to *-i-, while Baltic is presumed to have remodeled the paradigm on the basis of the weak suffix-form *-i- (01' *-'Ji-). Several variants of this analysis have been encountered in the preceding chapters. Schmid, as we have seen, posits an etymological relationship between the Balto-Slavic present and infinitive stems, and reconstructs a paradigm *mini!imi, *-iisi, *-iiti, pI. *-'Jimé, *-'Jité, *-'Jjénti (op. cit., 83). A similar formation, though with a diffeI'ent distribution ofstrong and weak forms, is assumed by W agner, 0- Verba, 52-3; both scholars compare the putative Germanic inflection *habaiP, 3 pI. *habjanp, and Wagner further adduces the Hittite type seen in 3 sg. lJ,alzai, 3 pI. lJ,alziyanzi. The same fundamental approach is embodied in Puhvel's reconstruction of an athematic type in *-eEY-j*-EY- (Lar. and the lE Vb., 53ff.): here the hypothesis that the syUabic form ofthe "palatal" laryngeal * EY yielded *-i­in Baltic obviates the need for assuming a long diphthongal suffix in Indo­European, but, as noted in § 12, leaves the quantity ofSI. *-i- unexplained. Other varieties of the ablaut theory make no attempt to establish a historical connection between *-~ and *-0-. A. Vaillant, Gram. comp. III, 437ff. operates with a Balto-Slavic type in *-ei-j*-i-, which he assumes to have been created in post-IE times through a complex series ofremodelings of earlier thematic and semithematic formations. Arecent study by W. Schmalstieg (Linguistique 8, 123ff. [1972]) posits a pre-IE element *-éje-j*-ejó-, from which evolved an lE ablauting suffix *-ei-j*-jo-, in Schmalstieg's view the immediate source ofthe stative conjugations ofthe attested languages.

These arguments have in large part been reviewed in ch. 3. Fundamental­ly, all suffer from the defect of assuming for Balto-Slavic and, with the exception ofVaillant's theory, for Indo-European, an athematic paradigm of a type whoIly unknoWll in the most archaic lE languages. To a greater 01'

les ser extent most depend also on the alleged formal parallelism between the Balto-Slavic statives and the Germanic third weak class; we have seen in § 59, however, that OS 3 sg. habad, -ed, pI. hebbiad and OE 3 sg. h'JefP, pI. habbap, the Germanic forms which most strongly invite comparison with the Balto-Slavic ~-presents, are in fact aImost certainly the result of a post­Germanic dialectal development. (On the Hittite type lJ,alzai, -iyanzi cf. § 54) 10). It may further be noted that reconstructions like Schmid's and

10) In a revision ofhis theory, Wagner, ZCP 25,161-73 (1956), compares the absence ofpalatalization in the 1,3 pI. and passive of 011'. saidid 'sits' (3 pI. sedait) and laigid 'lies' (3 pI. 'legat) with the alternation of -ai- and -a,- in Gothic . .But saigid

99

Page 50: Jasanoff.1978.Stative and Middle in Indo-European

Puhvel's, which presuppose a Balto-Slavic 1 sg. in *-mi, are badly compromised by the fact that the only finite' indicative form on which Baltic and Slavic agree exactly is the thematic 1 sg. *minjo ( > Lith. miniu, OCS mhñjq). While the possibility cannot in principIe be excluded that this form is simply the Balto-Slavic replacement of an earlier *mineimi 01'

*mineEYmi, it is clearlyunlikely that such a renewal ofthe 1 sg. could have taken place at a time when the putative alternation of strong and weak stems remained productive elsewhere in the paradigm.

A special case is presented by Cowgill's explanation of the e/~-class (Lg. 39,265-6), which combines aspects of both the ablaut and contrac­tion theories. As we have seen in § 55, Cowgill assumes an original ablauting type in *-ehrl*-hr , from which Germanic and Balto-Slavic allegedly created a series of thematic statives in *-hz-ie/o-. While in sorne respects more attractive than the solutions surveyed aboye, this reconstruction is as unconvincing in Balto-Slavic as in Germanic : in the last analysis it remains unclear whether the stative suffix *-e:. ever had a zero-grade *-hr , and doubtful that a medial laryngeal would have survived long enough in Balto-Slavic to participate in a development *-hzie- > *-'Jje- > *-ije- > *-1:­(> Balt. *-i-). Apparent counterexamples to the latter treatment, cited by Cowgill himself, are Lith. h'ia 'rows' from the se~ root of Olr. rlíi'd, 01 róa, etc., and OCS 3 sg. S'b-tuet'b 'will crush' (for *-ryet'b; cf. Tedesco, loe. cit.) from the root of Gk. 't'e;(pw, 't'pl')'t'6<;;, etc. Note also Lith. aria, OCS oryet'b 'plows' ( = Go. arjan) < *h2erhde/o-. Moreover, even if Cowgill's special assump­tions were otherwise admissible, his analysis would suffer from the same general weaknesses as the variants of the contraction theory discussed in the preceding section.

§ 85. A novel approach to the problem of the ~-presents is that of Kurylowicz, Infl. Gat. 39-54. Noting that certain statives, such as Lith. gari 'burns', OCS gorit'b ( : Gk. 8<ÍpofW;¡, etc.) and Lith. stóvi 'stands' (cf. Ved. pf. tastháu) can be taken to continue old perfects, Kurylowicz derives both. the Baltic third person form in *-i and the pre-Slavic 3 sg. in *-1: ( > -i-t'b) from *-ei, the lE 3 sg. perfect ending enlarged by the hic et nunc particle *i (cf. OCS védé '1 know' < *J!f!ida.-i). Accentualfacts aside, the development of Lith. -i and OCS -i from such an ending would have been completely regular ll

); Kurylowicz speculates further that on the basis ofthe 3sg. a new morphological type was created in which the reflex of final *-ei was generalized as a predesinential element throughout the paradigm. While he

'seeks' and the compounds of ic(c)- 'come', which show the same fluctuation, are not stative, and l seeno convincing reason to abandon Thurneysen's view (OlGr., 354) that forms of the type sedait have simply lost their palatalization by analogy with presenta of class B 1.

11) It would, however, be necessary to assume tha t the accen t of phonologically regular *mini was analogically retracted to the root.

100

"..,.. !

does not make clear the relative chronology of these events, it would probably be simplest to assume that the 3 sg. in *-ei was reinterpreted as a stem-formative in Balto-Slavic times, triggering the creation of 1 sg. *-ejo > *-jO) , 2 sg. *-ei-(s)ei, 1 pI. *-ei-me/o, etc. The development of the Slavic forms would then have proceeded normally, while in Lithuanian the shortening of *-ei to *-i in the third person would have led to the replacement of forms like 1 pI. *mineime, 2 pI. *mineite, etc., by the attested minime, minite I2

).

The ingenuity of this analysis is beyond question, but it is clearly untenable. As we shall see in § 90, the absorption of forms like 3 sg. *garei > *gl'hóre-i into the minifti-type appears actually to have been a comparative­ly late development which occurred at least partly independently in Baltic and Slavic. More decisive, however, is the simple fact, noted already in § 10, that in Old Prussian, where 3 p. turri, 1 pI. turrimai clearly show the same -i­as Lith. turi, turime, final *-ei is unambiguously retained (cf. dato sg. tebbei 'tibi', 2 sg. -sei = Lith. -8/', -síe-s; further examples are given by Stang, op. cit., 120) 13). Similarly, in those Latvian dialects which preserve the i­presents as a separate class, the 3 p. has a zero-ending (cf. gul' 'lies' = Lith. guli, cited by Stang, 320), pointing to a Common Baltic short vowel; original i-diphthongs are regularly retained in this position as -i. The possibilitl that the -i of Lith. mini continues earlier *-ei must thus be rejected 1 ).

§ 86. The history ofpast research into the ~-presents makes it clear, in my view, that attempts to explain Baltic *-i- as a phonologically regular shortening of earlier *-1:-, or to derive both *-i- and *-1:- from apophonic alternants of the same original suffix, are unlikely to prove fruitfuI. There remain, however, two theoretically possible ways of accounting for the discrepancy between Balt. *mini- and SI. *mhni- which have not been widely discussed. These are 1) to assume that the Slavic type in *-1:- is original, and that Baltic *-i- reflects the morphological influence of a distinct and etymologically unrelated formation; and, conversely, 2) to suppose that Balt. *-i- is original, and that pre-Slavic *-t'- was introduced into the stative paradigm under the influence of a historically separate type in which *-1:- was inherited. The fll'st assumption does not appear to lead to interesting results and will not be explored further here. The second possibility, however, is more promising.

12) Kurylowicz' explanation supersedes his earlier view (L'apoph., 128) that the Baltic and Slavic forms continue lE athematic formations in *-i- and *-i-.

13) l am indebted to Cowgill for having fIrst drawn my attention to the phonological weaknesses of Kurylowicz' theory.

14) Note also that ifthe paradigm ofthe r-presents had been created on the basis of the 3 sg. of the perfect, as Kurylowicz claims, it would be difficult to explain why the overwhelming majority of these forms show zero-grade, rather than o-grade of the root.

101

Page 51: Jasanoff.1978.Stative and Middle in Indo-European

As noted in passing in § 83, the Slavic present type in *-i- is not confmed to verbs like OCS mbnéti, but is equally characteristic of iterative­causatives like wsiti, pmsiti, etc.; to these may also be added the formally indistinguishable denominative type seen, e. g., in OCS gostiti 'lodge as a guest', preso goSéQ, gostisi, etc. < gOStE 'guest'. In the latter two categories it is difficult to escape the conclusion that *-i- is the result of a contraction : the only plausible source for the stem-vowel ofthe iterative-causatives, as Specht, op. cit., 78ff., recognized, is lE *-eje- while the gostiti-type is probably to be compared in part with Indo-Iranian denominatives like Ved. lcaviyáte 'acts like a sage' < lcávi- 'sage'. Note further that the circumflex intonation of *-i"- in these forms (cf. SO 3 sg: nosi and Vaillant, op. cit., 438) excludes the possibility of a derivation from an inherited long voweI. To be sure, a completely general development of *-eje-, vía *ije-, to *-i- may not be assumed: examples like trbje 'three (masc.)' show that originally disyllabic forms remain uncontracted in Slavic. In longer words, however, the only serious counterexample to the proposed rule is the nomo pI. in -bje ofmasculine i-stems (cf. pqtbje 'paths'). This ending may simply have been taken over from trbje, a development which would have been favored by the existence ofnom. pI. forms like synove 'sons' « *-e1/es) and consona.nt-stem forms like grazdane 'citizens' « *-es; cf. Sommer, Krit. Erl., 138); it is perhaps significant that contrary to the general Slavic pattern in masculine nouns, the expected nomo pI. *pqti would have been homophonous with the corresponding accusative form (PQti < *-ins). In the feminine, where the identity ofthe nominative and accusative plural was a recurrent feature, the phonologically regular ending was retained (cf. nomo pI. lcosti 'bones') 15). .

The Lithuanian i-stem plural in -ys (cf. alcfJs 'eyes' < alcis), taken togetherwith thenom. sg. in -f}s ofnouns with stem in *-ija- (§ 83), suggests that the contraction *-eje- (*-ije-) > *-i"- may date from Balto-Slavic, ra,ther than Slavic times; this question, however, will not be pursued fürther here I6

).

There is thus no reason to exclude out of hand the possibility that the stem vowel of the Slavic statives was originally short, and that it was lenghthened under the influence ofthe iterative-causatives and denomina-

15) 1 have benefited from useful discussions with Cowgill on this matter. It cannot be denied tha.t if there were an adequate explanation for the -i- of the iterative-causatives which did not require the assumption of a contraction *-e;je- < *-i-, it would be simpler to take the -i of kosti, rather than the -bje of pqtbje as analogical. (The feminine form tri 'three', of course, must be secondary in any event.) The important point, however, is that pqtbje does not constitute a telling argument against the proposed contraction.

16) Similarly, the infinitive in OCS -iti, Lith. -y ti may reflect a Common Balto­Slavic creation on the basis of a preterite « imperfect) 3 sg. in *-it < *-e;jet (cf. OCS 3 sg. rwsi, SC nos'i, with historically circumflex *-i-), just as the infmitive in BS *-eti

102

tives in *-iti, where *-i"- was inherited. The value of this hypothesis, of course, remains to be demonstrated; only if a direct lE source can be found for the Baltic stative type in *-1:- will it serve any purpose to suppose that the length of SI. *-i- is secondary. In the sections that follow we shall attempt to demon8trate that an explanation for the *-1:- ofBalt. *mini- i8 in fact available, and that a similar argument will account for the presence of *-i- in another Baltic categorywhose morphologicalhistory is obscure. The latter formation is the Letto-Lithuanian sigmatic future, the importance of which for our present discussion makes a brief excursus desirable.

§ 87. The East Baltic languages have an inflected future, the stem of which in Li.thuanian is characterized by -si- in the first and second persons, and by -s- in the third. The finite forms of the future of dúoti 'give' are thus as follows:

sg. 1 dúosiu 2 dúosi 3 d7l0S

duo dúosiva dúosita

pI. dúosime dúosite

The active participle associated with this paradigm exhibits a further morphophonemic peculiarity. Whereas in the i-presents the active partici­pIe is made by adding -nt- directly to the stem (cf. minint-, nomo sg. masco minjs), in the future the vowel -a- is regularly interposed after the temle sign, so that the future participle of dúoti i8 dúosiqs, dúosiant- rather than *dúosint- 17). This feature is alreadypresent in our earliest Lithuanian texts.

Historically, of course, the general affmities of this formation, which is almost exactly mirrored in Latvian, are with the sigmatic futures of Indo­Iránian, Greek, Italic, and Oeltic. The detailed history of the Bahic forms, however, is far from clear; in particular, the variable shape of the future marker, with its three distinct allomorphs, has no obvious parallels elsewhere in the family.

Apart from the synchronically isolated 3 p. duos, the Baltic future most closely resembles the Vedic (and, mutatis mutandis, Iranian) future in -sya-. The la;iter formation is fully thematic, and, accentuation aside, two memb6rs of the Vedic paradigm, viz., 1 sg. diisyámi « *diisyá; cf. Lith. dúosiu) and ptcp. diisyánt- (cf. Lith. dúosiant-) , can be exactly equated with their Lithuanian counterparts. These forms are in fact identified in most of

was presumably built from a preterite in 3 sg. *-et. The acute *-'i- of *-'iti would then be secondary. But it is perhaps more pro bable that *-iti originated in denominatives from i-stems, where participles in BS *-'ita- were directly descended from lE forms of the type found in Lat. mell'itus, Lith. akJJtas 'having eyes', etc. The extension of *-'iti to the iterative-causative class could then have taken place even before the contraction of *-iie- to *-'i-

17) Forms ofth; type dúo~int- do occur dialectally, but the synchronic isolation of dúosiant- makes it likely that this is the original formo

103

Page 52: Jasanoff.1978.Stative and Middle in Indo-European

the standard handbooks, where they are generally taken to continue a late 01' dialectal lE future in *-sjejo-. The rest of the Lithuanian paradigm, however, does not admit of so simple an interpretation, since late lE forms such as 1 pI. *dosjome 01' 2 pI. *dosjete would have yielded Lith. *dúosiame and *dúosiate, respectivelyI8). To circumvent this problem, Stang, op. cit. 379ff., has proposed a semithematic interpretation of the future suffix, according to which *-si- and *-sjo- would originally have had the same distribution as *-i- and *-jo- in the stative class. This solution, unfortunate­Iy, is open to many of the same objections as the arguments presented in § 82; furthermore, it is incapable of accounting for the 3 p. form d1los, which must therefore be explained apart from the rest of the futuI'e paradigm. Stang himself has convincingly shown that duiJs can only continue a pre­Baltic prototype *di5st, i. e., the 3 sg. of a simple sigmatic stem with no trace of a further formative *-i- or *-jejo_19). ./

§ 88. The very isolation of 3 p. d1loS would lead us to suspect that it is an archaism, and this impression is confirmed by further facts. In a large number of Lithuanian dialects (see Schmid, op. cit. 55ff. for a closer description) the dual and plural forms of the first and second persons are optionaIly 01' obligatorily athematic; representative examples from the dialect of Tverecius, cited by Otn;bski, Narzecze Twereckie 1, 372ff., are 1 pI. ve((sme, 2 pI. veaste beside "regular" veasim, ve"sit < Standard Litli. vedu, vesti 'lead'. Forms ofthis kind, as Schmid, op. cit. 58, has shown, are in aIl probability inheritances from Common Baltic, rather than reformations on the basis ofthe 3 p. in *-s. Likewise athematic in appearance is OPr. 2 sg. postasei 'wirst' (2 x), which is perhaps to be traceq. to an earlier *sta-s-sei; the meaning of this form, however, does not exclude the possibility of other analyses.

Athematic s-futures are also found in Celtic and Italic. The Old Irish sigmatic future is usuaIly reduplicated (cf. 3 sg. gigis 'will pray' < *gigedsti, memais 'will break' < *mimadsti) , thus recalling the Indo-Iranian desidera­tive; I'eduplication is lacking, however, in forms like 3 sg. reiss 'will run' < *retsti and seiss will sit' < *setst?: (Thurneysen, OJGr. 4,10f.), which differ from Lith. duos only in showing a primary ending. To these should doubtless be added Oscan and Umbrian futures like 3 sg. Osc. didest 'dabit', Osc., Umb. just 'erit', Umb. jerest 'feret', 3pl. Osc. censazet 'censebunt', Umb. jurent 'erunt'. The ltalic s-future, to be sure, is often taken to be thematic (e. g., by Buck, Gram. ojOsc. and Umb. 169), but against this view

18) Such forms occur only in Lithuanian dialeets which have given up i-inflection generally; see Schmalstieg, Slav. and E. Eur. Jour., 16, 120ff. (1958), for a short summary. Note that the 2 sg. dúosi can also be analyzed as a thematic form, though this is not necessary (cf. § 89).

19) On forms of the type dúosi, which occasionally appear in isolation, and are regular before the reflexive particle -s(i), see Stang, Vergl. Gram. d. balto Spr., 398-9.

104

speak 1 ) the regular preservation of the thematic vowel in presents of the third,conjugation (cf. Vest. didet 'dat', Marr.jeret 'fert'); 2) the patently athematic character of the Latin subjunctive type 1 sg. ,jaxim, axim, etc. (vs. secondarily thematized jaxo, capso, etc.); and 3) the form of the 3 pI. ending *-sent, which points directly to an athematic preform *-srfti, rather than to thematic *-sonti 20

).

J ust as such considerations make it probable that the athematic future is old in Baltic, the distribution offuture forms in the Rigveda suggests that the comparison of Balt. *-si- with III'. *-sya- is largely ilIusory. We have already noted that one of the two Lithuanian future forms with an exact correspondent in Indo-Iranian is the synchronicaIly exceptional participle dúosiant-,.which can be matcheii segment for segment with Ved. dasyánt-. The apparent etymological identity of these forms takes on added significance in view of the disproportionately high frequency of future participles in the Rigveda: Grassmann, Wb., lists twenty-nine participles, representing twelve roots, compared with only twenty finite forms, representing ten roots. This pattern, which differs notably from that displayed by ya-presents of the normal (classIV) type, would seem to indicate that finite forms like dasyát'i are historical back-formations from participles like dasyánt- ; the model for their creation, of course, would have been supplied by such regular pairs as pú,¡yati :púllyant-, ásyaU: ásyant-, etc,

It thus appears that a future participle in *-sjont- can legitimately be reconstructed for dialectal Indo-European; note further the isolated OCS participle bys~st-jbysqst- 'about to be, future', the latter variant of which (cf. also OCz. probys7lcny 'useful') is doubtless tobe referred, along with Lith. b1Ísiant- and Av. busiiant~ 'id.' (Ved. bhavillyánt- is a younger form) to an inherited *bhusjont_ 21 ). On the basis of such forms a complete finite paradigm in *-sya- was evidently created in Indo-Iranian, while in pre­Baltic, where no comparable renewal took place, athematic forms of the type 1 sg. *dosm'i, 1 pI. *dosme, 2 pI. *doste, etc., were for a time preserved intacto It is in this historical setting that the later creation ofLith. dÚOSÍ7l, dúosime, dúosite, etc. must be viewed, and not in the false light of the equation dúosiu = dasyámi: the identity ofLith. -siu and Ved. -syámi is, in Meillet's phrase, a "mirage de la grammaire comparée."

§ 89. In my view, which is essentiaIly the same as that of Schmalstieg, Slav. and E. Eur. Jmlr.16, 120ff. (1958), the si-forms ofthe Baltic future

20) The future, in faet, is most pro bably the category whieh provided the model for the replacement of *-ont by *-em in Ose. 3 pI. preso fiiet 'fiunt', staiet 'stant'; it is difficult to follow Buck in attributing *-ent and *-sent to the influenee of isolated athematic forms like sem 'sunt'.

21) The close agreement of Lith. b1Ísiant- and Av. busiiam- makes it unlikely, in my judgment, that OCS bysqst- was created within Slavic, as claimed by Aitzetmüller, Gedenkschr. Brandenstein, 11ff.

105

Page 53: Jasanoff.1978.Stative and Middle in Indo-European

can be explained entirely on the basis of an earlier athematic paradigm in *-S-. The process by which the future sign carne to acquire a following *-i- is most clearly revealed by the exceptional paradigm of the ver b 'know' in our oldest attested Baltic language, Old Prussian. In the singular, OPr. waist has the athematic present forms 1 waidmai, 2 waisei, and 3 waist; historical­ly, this inflection has replaced that of an earlier perfect, the 1 sg. of which survives in 008 védé (OPr. -mai = pf. *-ai x preso *-mi). We should expect the plural to be athematic as well; the actually attested forms, however, are 1 pI. waidimai and 2 pI. waiditi, with an etymologically extraneous *-i-. The origin of this vowel is no doubt to be sought in the fact that before the Oommon Baltic loss of the singular: plural opposition inthe third person, the 3 pI. of waist would have been *waidint(i) 01' *waidir, with an ending l'epresenting an earlier perfect 3 pI. in *-1Ji 01' *-1'. A form like 3 pI. *waidint(i) would naturally have been subject to renalysis as stem *waidi­+ 3 pI. marker *-nt(i), and such a reinterpretation could in turn have led to the generalízation of -i- to the remaining persons of the plural. Typological­ly, developments of this kind are not unknown outside Baltic : a precisely similar case is that of the plural preterite endings 1 *-um and 2 *-up in Germanic (cf. also Go. 1 duo -u < *-u-we, 2du. -uts < *-u-tes), the *-u- of which has been extended from the regular 3 pI., in *-un « *-tl-t). G. Klingenschmitt (personal communication) has called my attentioll to the fact that the-a- ofthe Armenian passive aorist (type beray, berar, etc. 'I was carried, you were carried') can plausibly be traced to the 3 pI., where lE *-tl-to yielded -ano Note also the replacement of -ci- by -mx- in the plural and dual of the sigmatic aorist in Greek, where -IX- is etymological onll in the 3 pI. in -crIXV ( < *-8t1-t + analogical -v) and 1 pI. in -lXfLe:v (*-~men) 2 ).

8imilarly, the inflection of the Baltic future can be accounted for by assuming that the pre-Baltic 3 sg. *dost originally corresponded to a 3 pI. *dósint < *dóstl-t. From such a form it would have been a simple matter for *-i- to be extended to the rest of the plural and dual paradigm, thus giving rise to the immediate antecedents of Lith. 1 pI. d1losime, 2 pI. dúosite, etc. The new forms did not at the outset completely replace the old : a slight but perceptible functional contrast between the types dúosime and dúosme is maintained in sorne Lithuanian dialects to the present day (cf. Otrftbski, loe. cit.; Fraenkel, Malcher Pietkiewicz, 34f.). The singular paradigm was largely unaffected by these developments. Here Lith. duíJs directly continues *dast, and 2 sg. dúosi is most simply taken as *duos-sei, i. e., as a straightforward athematic form with the substitution ofBalt. *-sei for *-si (cf. § 42). Only in the first person was *dósmi replaced by dúosiu, a renewal which was favored both by the form of the present participle and by the regular Baltic correlation ofplural and dual forms in *-i- with 1 sg. forms in *-jo. ~~~-

22) The -IX ofthe 1 sg., of course, is also regular « *-81/1); since it is synchronicalIy a desinence, however, it is unlikely to have had much effect on the generalization of -cr(J,- at the expense of -cr- in the dual and plural.

106

... It is impossible to give an exact chronology for these remodelings, save,

oí course, to note that the extension of *-i- from the 3 pI. could only have occurred before the loss of the 3 pI. as a distinct form in Oommon Baltic. Slavic preserves no unambiguous traces of a future in *-si- 01' *-s1:-. The participial fOl'm bys~st-, which could in principIe represent a replacement oí bysQst- « *bhusjont-) under the influence of a finite paradigm *bysQ, *bysisi, etc., can equally weIl be explained, as Hollifield has pointed out to me, as the reflex of an earlier form in *-8t1-t-: it is significant that the s-aorist participles of Avestan, which are historically ofthis shape, frequently have future value (cf. Reichelt, Awest. El.,328). Nevertheless, the only cel'tain terminus post quem fOl' the substitution of *-si- for *-s- in the plural and dual of the future is the Balto-Slavic sound change of *t1 to *in; there exists at least a theoretical possibility that forms of the type *dosime and *dosite once existed not only in Baltic, but in pre-8lavic as welI.

§ 90. Baltic examples of the extension of i-inflection from an original 3 pI. in *-int( i) are nQt confined to OPr. waist and the Letto-Lithuanian future; it is significant, moreover, that traces of a similarprocess can be o bserved in Slavic. OPr. 1 pI. waidimai and 2 pI. waiditi invite immediate compal'ison with Lith. stoveZi 'stand' (1 pI. stóvime) anddéveli 'wear' (1 pI. devime) , which likewise continue original perfects (cf. §85; fOl' déveU cf. Ved. dadháu)23); inseparable from these is 008 bojati s~ 'fear' « *-éti), which belongs with the family of Ved. bibMya and Gmc. *bibaip (§ 62).

Here too, in the usual view, belong Lith. gareli, OOS goréti, -it'b 'burn', OOSpoléti, -it'b 'burn', and boléti, -it'b 'be ill' (Stang, Slav. U. balto Vb., 24). To be sure, the o-grade of these forms is ambiguous: in the case of BS *gar¿ti in particular, an earlier athematic present *gl'hor-j*gl'her-, rather than a perfect, is suggested by the vocalism and morphology of Gk. 8ÉpofLlX¡ (cf. §§ 38, 70, 72). But since a zero-grade 3 pI. ending would have been regular both in the o-grade present and in the perfect, it is not important to make a distinction between the two categories as far as Balto-Slavic is con­cerned24

).

A second group of verbs whose membership in the f,-class is secondary consists of original acrostatic presents with *e: *e apophony (cf. § 38). This type, which likewise ended in *-tl-ti in the 3 pI., was formerly well­represented in Balto-Slavic: directly attested examples include OLith. esti 'eat', 3 p. esti, OOS jasti 'id.', 3 sg. jast'b; OLith. be'gti run', 3 p. be'gti, OOS Miati « *-éti) 'id.' (cf. note 26 below); perhaps also OLith. sérgeti 'protect',

23) OtherwiseBammesberger, Lg. 50, 687-95 (1974); Idonotfindhis arguments persuasive, although the original function and distribution of *-')!- in these forms remains problematic.

24) Nor, assuming that the acrostatic analysis is correct for sorne 01' alI of these verbs, is it possible to decide with certainty whether the pre-Balto-Slavic forms were middle, like pre-Gmc. 3 sg. *torpór and *konkór, 01' active, like pre-Gmc. *kónke. The latter alternative seems likelier (cf. note 35).

107

Page 54: Jasanoff.1978.Stative and Middle in Indo-European

3 p. sérgti ( < *strgti). Sevveral ve~b~, .ho,we¡er; ?ave replaced forms of thís kind by paradigms in *-'i-. OCS vtdet~, -~t'b see lS perhaps best taken as the replacement of an earlier athematíc *yAidmi, *-si, *-ti, the optatíve of which survíves, at least indírectly, in OCS ímpv. 2, 3 sg. viidh 25 ). Both athematic and i-forms of the root *yeid- are found in Baltíc also: for the former compare Líth. veizdéli 'look' (OLíth. 1 sg. veizdmi)26), for the latter Lith. pavydéli 'envy' (cf. Lat. inuidere), 3 p. pav{¡di. Despíte Kurylowicz, L'apoph.,292, the acute intonatíon of BS *tddf,- and íts synchroníc assocíatíon wíth an athematíc paradígm in both branches make it doubtful that thís stem was created by secondary lengthening from a statíve of the usual type (cf. § 95). Rather, it appears that the startíng point for the Baltic and Slavíc stem-forms in *-I:- was a 3 pI. *yUinti < *-1Jti, replacing an earlíer 3 pI. *y,eid1Jii; the new morphologícal zero-grade *y,'id- would have stood in the same relatíonship to *yZid- as díd *-i- to *-ei- in athematic presents of the tradítíonally reconstructed type. For the ablaut compare the parallel Baltíc paírs Líth. níedéti 'shun' vs. Latv. nfdet (besíde nfst) 'scorn' and Latv. ziédét 'bloom' vs. Líth. iydéli, 3 p. i{¡di 27 ).

As in the case of the future, ít ís dífficult to establísh a relíable chronologícal interpretatíon of these facts. The plural of the ver b 'know' in Slavíc (cf. OCS 1 vém'b, 2véste, 3 védfit'b) seems to show that the extractíon of the stem *y,aidi- from the 3 pI. was a purely Baltíc development. The equatíons Líth. gari = OCS gorit'b and Líth. pav{¡di = OCS vidit'b, on the other hand, indícate that here the generalízatíon of i-forms may have begun in the common período It ís possíble, for example, that the Balto­Slavíc paradígms of these verbs were already characterized by *-i- in the plural and dual, though not in the singular: both branches would then have completed the extensíon of *-i- independently, Slavíc innovating further by lengthening *-i- to *-'i- (cf. § 86). That the dístribution of*y,'idi- was still partly restrícted in Balto-Slavíc ís suggested by the athematíc forms cíted above; for *gari- note the OCS preso ptcp. gorqst- (beside regular gorfist-),

25) On the oas athematic imperative see note 43 below. Although the acrostatic type in *e: *e appears to have been associated with aoristic roots from lE times, *'J!Éidrni itself need not have been inherited from the parent language; see also § 97.

26) The -zd- of these forms, which is confmed to Baltic, may have arisen in the imperative, where *Jiéid-dhi 'look!' would have given *yeizdi, 01' (less likely) in the 3 sg., where a stem *yeizd- could have been extracted from *yeisti < *Jiéid-ti.

27) The latter verb, referred by Pokorny, IEW 355, to the root *{jei-, is offurther interest in showing that the root enlargement *-d- (*-dh-) could form athematio presents in Baltic; it is thus probable that the i-inflection of the Lithuanian iterative type in -d- (cf. rnérdéti, 3 p. rnérdi) is likewise due to the original presence of a 3 pI. in *-inti. Other t-presents of aorostatio origin include dialectal Slavic béii­beside béiati (Vaillant, op. cit., 382f.), and Latv. 1 pI. raUdirn, 2 pI. raUdit « raUdát 'weep'), the correct explanation of which was seen already by Endzelin, Filologu biedr'ibas raksti 8, 107.

108

which can only date from a tíme when the incorporatíon of goreti into the normal stative type was not yet complete28

).

§ 91. The relevance of the formatíons díscussed in the preceding seotíons to the problem of the statíve presents should now be olear. The hístory of the future, perfeot and aorostatic present types in Baltío and Slavío undersoores the fact that the 3 pI., as the non-singular verbal form par excellence, is typologícally well-suíted to serve as the point of departure for formal innovatíons elsewhere in the plural and dual paradígm. It ís true that many, and conceívably even all of the remodelings studíed aboye belong to the post-Balto-Slavic períod, when f-presents were already avaílable to províde a model for the extraction ofnew stem-forms in *-f,­from the 3 pI.; neveI"theless, nothing whích we have assumed for the development of the three categoríes in questíon would have been ímpossible or unnatural íf the statíve type had not yet been present in the language. (Compare the Germaníc example cíted in § 89, where *-u- was able to spread from the 3pI. despíte the absence of a pre-exísting type in *-um, *-up, *-un elsewhere in the verbal system.) In short, the possíbílíty cannot be excluded a priori that the *-f,- of the Balto-Slavic statives ís ítself ultímately attríbutable to a process of generalízatíon from the 3 pI. The ímplícatíons of thís fact wíl1 be explored in detaíl below.

§ 92. We have seen in §§ 82-5 that attempts to relate the stem-vowelof Balt. *mini- and SI. *mhn?:- to the lE suffíxes tradítíonally reconstructed *-jejo-, *-ejejo- and *-e- have not met with notable success. As an alternatíve approach, let us now consíder the possíbílíty that the present of BS *mintti ís to be explained on the basís of the same formatíon as íts Germanic counterpart, the third class weak verb *munan, *-aiP. In § 68 ít was argued that Gmc. 3 sg. *munaiP ís best regarded as a transformatíon of an lE perfect míddle * (me )mnór, the hístorícally regular outcome of whích, *munai, was suffíxed further by the *-p of the 3 sg. actíve. Since Balto­Slavíc, like Germaníc, appears generally to have elíminated reduplícatíon from íts survíving reflexes ofthe lE perfect (cf. Líth. stóvi, OLíth. liekmi '1 remain', OCS boit'b, etc.) the pre-Balto-Slavíc paradígm corresponding to that of pre-Gmc. *munai would presumably have had the following form (cf. §§ 65, 78):

sg. 1 2 3

*mina'i *mintai *minai

( = pre-Gmc. *munai) ( = pre-Gmc. *mundai) ( = pre-Gmc. *munai)

28) Note also the deviant oas participle vidorn'b. Further evidence for "mixed" paradigms in Baltic is parhaps to be seen in 1 sg. forms such as Lith. stóvrn'i, -vf¡drni, if¡drni, rnérdrni; these are, however, late (Ruhig) and open to other possible in terpretations.

109

Page 55: Jasanoff.1978.Stative and Middle in Indo-European

pI. 1 2 3

*minm_ 29

*mind_ 29

*minintai

( = pre-Gmc. *mnnm-) ( = pre-Gmc. *mnnd-) ( = pre-Gmc. *mnnnndai)

Two preliminary remarks are in order. The aboye reconstruction presupposes that Balto-Slavic originally had primary middle endings ofthe type in *-i, like Greek, Indo-lranian and Germanic, rather than ofthe type in *-1', like Anatolian, Italic, Oeltic and Tocharian. This assumption will play no role in the remarks that follow: final *-ai may everywhere be replaced by *-ar without significant consequences for the discussion as a whole. Less trivially, however, we shall also assume in what follows that pre-Balto-Slavic replaced the inherited 3 pI. in *-rai (= Ved. -re) by *-intai < *-1}toi; although this supposition cannot independently be con­firmed, comparable renewals of lE *-ro(i) are found in every lE language in which the middle has survived as an autonomous category30).

§ 93. As the initial step in the remodeling of the perfect middle in Balto­Slavic we may speculate that, as in the typologically similar cases cited in §§ 89-90, the *-i- ofthe 3 pI. wasgeneralized to theremainingpluralforms, yielding 1 pI. *minim- and 2 pI. *minid-; the same process would presum­ably have introduced *-i- into the endings ofthe dual as well. Oonsider now the effect on this paradigm of the general Balto-Slavic elimination of the middle desinences. In the plural and dual, where the old middle forms contained a union vowel *-i- before the terminations proper, the simplest course would be to assume that this vowel was preserved intact, while active endings were substituted for their middle counterparts. 3 pI. *minintai would thus have been remade to *mininti, while the original forms of the 1 pI. and 2 pI. would have become *minime and *minite, respectively. Outside the singular, therefore, the original perfect middle paradigm would have yielded forms essentially identical to their attested equivalents in Baltic.

It is less easy to predict how the loss of the middle would have affected the paradigm of the singular. The 3 sg. middle *minai could in principIe have had several developments. As in Germanic, the obsolete ending *-ai might have been reinterpreted as part of the stem: in this case *minai would probably have been retained, since the 3 sg. in Balto-Slavic is typically characterized by a zero-desinence (cf. below). Alternatively, *-ai could have been replaced by its athematic active analogue *-ti, yielding a new 3 sg. *minti 01', with generalization of *-i- from the plural, *miniti.

29) As in the corresponding Germanic forms, we shaH not attempt a reconstruc­tion of the 1 pI. and 2pI., middle endings.

30) Thus, only Indo-Iranian and Tocharían preserve 1'-endings in the 3 pI. middle at aH, the latter most clearly in the isolated form B stare 'are' < *sth2-ró (for *sUfre, presumably because it was enclitic; cf. 3 sg. ste 'is' < *sth2-ó). Even in Indo-Iranian, *-atai and *ata appear beside *-rai and *-m[ n].

110

T I

N either of these possibilities seems in fact to have been exploited. Instead, Balto-Slavic evidently extended to the nascent i-class the pattern of the thematic conjugation, where the 3 sg. was formally identical to the stem­allomorph which appeared before the endings of the 1, 2 pI. and 1-3 duo 3 sg. *minai was thus replaced by *mini, which stood in the same relation to 1 pI. *minime, 2 pI. *minite, etc. as did, e. g., 3 sg. *yede to 1 pI. *yedeme, 2 pI. *yedete, etc. 31 ).(If the change of *-eje- to *-f- had already taken place (cf. § 86), the inflectional type ancestral to OOS 3 sg. nosi(t'b), 1 pI. nosim'b, 2 pI. nosite would have conformed to this pattern as well.) An additional factor in the choice of *mini as the new 3 sg. may have been the following. In the 3 pI., the replacement of *minintai by *mininti was presumab~y acco~:par: nied by a period of fluctuation in the use of these forms, wlth *m~mnt~ gradually gaining the upper hand. To a speaker ofpre-Balto-Slavic at this stage fmal *-ai and *-i would have appeared to be in free variation in the 3 pI.; it is not entirely improbable that this situation could have induced a similar alternation in the 3 sg., where the inherited ending *-ai would thereby have acquired a variant *-i.

With *mini established in the 3 sg., the history of the remaining singular forms would for all practical purposes have been determined. The original 1 sg. *minai would almost surely have been supplanted by an active form in *-001' *-jo, just as in Germanic 1 sg. *mnnai was remade to *mnno (cf. § 78); in Balto-Slavic the presence of an apparent stem *mini- elsewhere in the paradigm would obviously have favored the selection of *minjo, rather than *mino, as thenewform 32). In the 2 sg. thereconstruction ofthe Balto­Slavic prototype is unclear. Probably the simplest assumption would be that a new 2 sg. was created according to a proportion of the type 3 sg. *yede: 2 sg. *yedei:: 3 sg. *mini: 2 sg. X, X = *mini + i> *mini; it would then be possible to take BS *-i as the direct source of dialectal Lith. -i, reflexive -ts (cf. Stang, op. cit. 409)33). Rere too may belong OPr. 2 sg. tnrri, if this form has not simply been generalized from the third person; Standard Lithuanian -í(e) (= Latv. -i(e)) is probably best regarded as an analogical creation on the model of the regular thematic type in *-ja-. In Slavic a 2 sg. in *-i perhaps underlies the attested form in -isi (OOS), with

31) Of. § 42, where these forms have been discussed. I follow Watkins, Idg. Gr.1I1. 1, 219, in taking OOS -t'b to be a secondary accretion to the 3 sg. ending, possibly identical with the demonstrative pronoun t'b 'ille'.

32) It is not inconceivable that the choice of *minjO was also influenced by the continued existence at this stage of Balto-Slavic of a synonymous -je(o- presEffit cognate with Gk. flcdvofl<X~ and Ved. mányate. The 1 sg. could also, of course, have ended in *-ijO (i. e., *-i-i5), from which *-jOin both branches would have been regular in any case.

33) With the acute intonation, historically pro blematic, of other Lithuanian 2 sg. forms.

111

Page 56: Jasanoff.1978.Stative and Middle in Indo-European

*-sei suffixed directly to the inherited ending 34). If of this origin, 2 sg.

mbnisi, etc., may have played a role in the replacement of*-t- by *-i"- within the subsequent history of the stative paradigm in Slavic (see below) 35).

§ 94. The fmal result of these developments would have been the creation of an indicative paradigm which may be taken to represent the present of *min~ti in Balto-Slavic:

sg. 1 2 3

*minío *mini" *míní

pI. 1 2 3

*míníme *míníte *min~nti

These forms, with the obvious exception of the 3pI., were preserved unchanged in Baltic. In Slavic, on the other hand, *-t- was apparently replaced by *-i"- under the influence of the iterative-causatives and \ denominatives in *-i"- (cf. § 86); themergerofthetwo types, itwill benoted, would have been phonologically regular in the 1 sg. (cf. OCS nosq < *-jO < *-íio < *-eiOJ and probably in the 2 sg. as well (cf. OCS nosisi < *nosí + si, with 2 sg. *nosi < *-ejei) 36). In the 3pI., contrary to the pattern elsewhere, the stative ending *-intí, synchronicallyreinterpreted as *-i"nti, superseded the iterative-causative 3 pI. in *-ionti < *-ejonti; OCS 3 pI. nos~t'b thus represents an analogical replacement of inherited *nosot'b. For a discussion of the corresponding modal and participial forms see '§ 98.

It emerges, then, that the presents of Lith. minéli and OCS mbnéti can both be explained as morphological transformations of the lE paradigm posited in ch. 3 as the prototype of the Germanic classIII weak present *muno, *-ais, etc. It remains to be seen, however, whether the lexical membership of the minÜi-class favors such a derivation for the Balto­Slavic stative type as a whole.

34) Alternatively, we could explain both the Slavic and Standard Lithuanian forms on the basis of a thematic (*-jejo-) 2 sg. in BS *-ei < *-jei. The analogical process which created the dialectal Lithuanian 2 sg. in -'i would then have been of post-Baltic date.

35) The above account presupposes that *-i- was generalized from the 3 pI. before the 10s8 of middle inflection. If we assumed that these developmen ts had occurred in the opposit~ ?rder, it would be less easy to motivate the creation of 1 sg. *minjO and 3 sg. m~n~; we would also be unable to explain why the extension of *-i- from 3 pI. *-inti was general in the stative class, but only sporadic in other categories.

The possibility of explaining the -i- of the Baltic sta,tive class as an analogical extension from the 3 pI. was already seen, though not fully exploited, by Schmalstieg, Int. Joum. oi Slav. Ling. and Poet.1, 181-3 (1959).

36) The existence of 3 sg. forms of the type *gori < *-ei, whatever their origin, may have played a role in the substitution of *-'i- for *-i- in Slavic. Typologically, the merger of the two i-conjugations recaIls, e. g., the substitution of It. capire fol' Lat. capere (1 sg. capia¡.

112

......

I I

§ 95. As in Germanic, the expansion of the perfect middle in pre-Balto­Slavic would have been most likely to affect roots with inherited perfect actives or inherited middle root aorists (cf. §§71-2). Examples of both types are not difficult to find:

Lith. budéli, bUdi, OCS b'bdéti, -it'b 'be awake': an active perfect is presupposed by Ved. (subj.) búbodhati. Note also the middleroot aorist 3 pI. ábudhran, abudhmm, inj. budhánta (with -anta for -ran; cf. §35), ptcp. budhanár; Gl\:. E7tó()e;-ro is doubtless a thematization of a similar formo Both Vedic and Greek, moreover, show the corresponding perfect middle: cf. Ved. ptcp. bubudhaná- (with post-Vedic finite forms), Gl\:. 7tÉrcUcrTGÜ, plpf. 7tÉ7tUcrTO.

OCS dr'b~ati, -it'b 'hold': the Vedic classIV present drhyati 'is firm' and perfect middle participle dailrha1Jár suggest the former presence of a middle root aorist, a trace of which evidently survives in YAv. d<J'l'<Jzano. p'J'l''Jea­V.3.41 '(exacting) strict (d'Jr'Jzana-) penance'. These forms point to root­final *-gh; a variant in *-gh is attested in YA v . drazaite 'holds', which in this respect more nearly resembles the Slavic formo

Lith. guléli, guli 'lie' : as suggested in § 33, an etymological connection is possible with one or both of Gk. ~A'~TO, ~É~A'I)TIX~ and Toch. A kulatii'l' , B kuletiir (classIII) 'slackens', the latter verb, of course, being the normal Tocharian continuant of a middle root adrist.

OCS lezati, -ít'b 'lie' : the root *legh- attests a perfect in Gl\:. AEAOXU!.IX· Aé:XW

Yé:vofLÉv'Y) and perhaps in Hitt. laki 'makes crooked'. The corresponding middle root aorist survives in Gl\:. AÉXTO.

OCS p1'~-lbpéti, -i~b 'be stuck (to)': Toc~. B lipet~'l' 'remains over' implies an lE aOflst 3 sg. *hpó. Under the analysls offered ID the preceding chapter Gmc. *liban, *-aip 'live' would represent an exact parallel to the Slavic form (cf. § 72).

Lith. §Vytéli, Suyti 'glitter', OCS 8Vbtéti s~, -it'b 'shine' : the long root vowel ofthe Lithuanian form reflects the morphologicallengthening, common in verbs ofthis class, described by Kurylowicz, L'apoph. 291f. (Lith. §Vitéli, 1 sg. Suitu, from which Suytéli is derived, has abandoned i- for a-inflection.) An aorist middle participle Svitanár is found once in the Rigveda (6.6.2).

Lit~. ty'léli, tylí 'be silent': as in the preceding item, the long root vocalIsm lS secondary. The relationship of tyléli to Gmc. *pulan, *-aip 'suffer, endure' is the same as that ofOCS -lbpéti to Gmc. *liban, *-aip (cf. above). As noted in §72, Gk. perf. 1 pI. TÉTAlXfLé:V (cf. OLat. tetuli") and aor. 3 sg. ~TA'a point to the existence of an active perfect and root aorist in Indo­European.

OCS tr'bpéti, -it'b 'suffer, endure': the original existence of an active perfect is shown by Go. pa'l'f 'needs' and related Germanic forms. Gk. subj. 1 pI. TlXp7tWfLé:81X is probably thematized from an earlier root aorist *trp(t)ó; a perfect middle participle tattpanár is found RV 10. 95. 16. It is also possible that the Slavic verb continues an original present of the type assumed in § 38 and § 70 for Toch. A tsa'l'watiir and Gmc. *parbaip; the attested zero-

113

Page 57: Jasanoff.1978.Stative and Middle in Indo-European

grade could then have been extended from the stem in -é-, where this vocalism was regular.

§ 96. Here also belong two Slavic verbs which are exceptional in having infinitives in -ati rather than -éti:

OOS shcati, shcit'h 'mingere': an lE middle root aorist is indicated by Toch. A 3 sg. *sikata1' 'overflows' and, as noted in § 33, Ved. subj. 3 sg. sécate. The sense of shCati makes it unlikely that this verb continues an earlier perfect middle. Instead, it seems probable that the present stem shci- was extracted from an athematic aorist present of the type found in the Tocharían third class and in Gmc. *fulgan, *-aiP (cf. § 66); from a morphological point ofview, of course, the loss ofreduplication in Balto­Slavic would have rendered such forms indistinguishable from perfect middles. If correctly explained in this way, the presents of shcati and Toch. A sik- would constitute a genuine word equation, permitting the reconstruction of an lE 3 sg. preso *sikór.

OOS s'hpati, s'hpit'h 'sleep': the immediate comparandum is Hitt. 3 sg. 8uppari, suptari 'sleeps', the historical status ofwhich, however, is itselfnot entirely clear. If, as seems not impossible, 8uppari is simply the middle corresponding to Ved. preso 3 sg. svapiti (cf. also impv. 3 sg. sváptu (AV)). the most direct explanation of SI. *s'hpi- would be on the basis of a middle root present of the same type. It is perhaps likelier, however, that the Híttite present is a secondary creation, formed by adding primary endings to an older root aorist *sup(t)ó (cf. martari 'dies' beside Ved. aor. inj. mrta); in this case the Slavic present would invite typological comparison with the Vedic perfect middle participle su?upaná- and inj. 2 sg. su?upthalp (Brahma­nas).

The verbs shcati and s'hpati are of additional interest in that they strengthen the assumption, implicitly adopted here, that the Balto-Slavic present suffix *-t- is historically independent of the infinitive stem in *_e_ 3

\ The association of shci- and s'hpi- with extra-present forms in -a­strikingly recalls the association of class III (IV) presents in Tocharian with subjunctives and preterites in OToch. *-a- (§ 14).

§ 97. In a number of cases the interpretation of a Baltic 01' Slavic stative under the perfect middle hypothesis is uncertain. If, as suggested to me by Schindler (personal communication), the -e:- ofLat. sed'í and Go. 1 pI. setum 'we sat' reflects a sound change -Vzd->-Vd- of lE date (cf. Av. hitJaiti = Ved. stdati), it would be possible to derive Lith. sedeli, se'di and

37) To be sure, it would still be possible under Cowgill's theory of the t-presents (§84) to derive shéi- and s'hpi- from *sik-hdejo- and *sup-hdejo-; presumably *-hdejo- and *-hriejo- would have yielded the same result in Slavic. Note that whatever its true explanation, shéi-, which is not semantically stative, is unlikelyto be a mere Entgleisung for earlier *sik-jejo-.

114

OOS sedeti, -ith from a reduplicated perfect middle stem *sed-<*sezd-38).

On the other hand, it is difficult to separate the vocalism ofthis verb from that of the OOS thematic aorist séde 'sat down' 01' from that of OLith. 3 p. se:~ti 'sits down', which points to an earlier acrostatic present in *e: *e. Note also the short root-vocalism oí OOz. sedeti. An unambiguous determina­tion of the status of BS *seditti does not for the moment seem feasible 39

).

BS *~'ídeti, as we have seen above (§ 90), is probably best separated from Gmc. *witan, *-aiP and explained without reference to an inherited perfect middle. It should be borne in mind, however, that if our interpretation of the root-form *~'íd- as a morphological zero-grade is correct, the new allomorph *~'íd- could in principIe have replaced the phonologically regular zero-grade *y,id- in any and all grammatical categories where the latter was inherited. Thus, even if Balto-Slavic had originally possessed a series of forms precisely analogous to Gmc. 3 sg. *witai(p) and Ved. vidé, such forros would no longer necessarily be uniquely recoverable.

Not all t-presents, of course, are directly referrable to athematic forroations. OOS stojati, -it'h 'stand' is most naturally explained as an old jej 0-presen t * sthz-ie j 0- which, following the deca y ofthe in transitive type in *-jejo- in Balto-Slavic, was secondarily transferred to the t-class. The oes present is thus probably the replacement of an earlier *stojq, *-jesi, which may be compared directly with Hitt. ti~a- 'step' and Gmc. 3 sg. *sta(j)ip (OHG stet) , 3pl. *stajanp (OHG stant)4 ). Another likely transfer of this kind is OOS kypéti, -it'h 'boil', for which Skt. kupyati (Ep.) 'be angry' and Lat. cupw guarantee an old jejo-present, but for which no clear evidence of an inherited perfect 01' root aorist can be cited. Lith. syjeli, sfJji 'be connected' is doubtless to be compared with Ved. áva syati 'binds' and Hitt. iBIJai-; the precise interrelationships ofthese forms, however, are unclear (cf. §54).

§ 98. The examples discussed in §§ 92~ show that a substantial number oH-presents can plausibly be regarded as morphological transformations of earlier athematic middle formations. One of the most conspicuous advan­tages ofthis analysis is that it accolillts directly for the Balto-Slavic 3 pI. in *-inti, which must otherwise be explained by a weak analogical argument (cf. § 82). As we shall now see, the theory presented above also permits a straightforward interpretation of the present participle in *-int- and optative in *-'í-.

3M) ~uch forms as Av. pf. opto 3sg. hazdiiiitand Gk. 'l~ú) would then, of course, have to be explained as analogicall'emodelings.

39) 1 sg. se'dmi 'sedeo' (R uhig) is eviden tly a form of the type stóvmi (cf. note 28). 40) It is not improbable that the replacement of *stoje- by stoi- was triggered by

the prior creation of the stative infinitive *stojeti, On ORG stet, stiint, see Cowgill, Journ. of lE St7¿d, , 1,298-9 (1973), corl'ecting my earlier attempt to explain these forms as tl'ansformedmiddles (NSF ReportRARV-LING-01-72, 291-307 [1972]).

115

Page 58: Jasanoff.1978.Stative and Middle in Indo-European

In §§ 75-6 it was suggested that Germanic third class present participles such as Go. munands are best explained as apophonically normalized reflexes ofthe lE participial type in *-ent-/*-r!J-, rather than as entirelynew analogical creations on the basis of the class III finite paradigm. The corresponding Balto-Slavic forms (cf. Lith. minint(i)-, OCS mhn~§t-) are equally amenable to such an analysis: here it is only necessary to assume that the zero-grade suffix-form *-int- < *-r¡J;- was generalized under the influence of the present stem in *-i-. In Slavic *-int- was extended to all presents in *-í"-, so that phonologicallyregular OCS *nosQst- was replaced by nos~st- in the same way that *nosQt'h was superseded by nos~t'h in the 3 pI. (cf. §94)41). . .

Likewise exhibiting an athematic structure are the inherited forms ofthe optative. In Slavic, where the original situation is relatively clear, the plural and dual forms of the imperative ( = lE optative) are characterized by -i- + the personal endings, and are thus indistinguishable at the segmentallevel from the corresponding indicative forms. 'fhey differ from the indicative, however, in showing acute intonation of the suffix (cf. Vaillant, op. cit., 528), thus recalling athematie imperatives like OCS 2 pI. dadite 'give' « *dadUe), in which the mood sign -i- is simply the regular Slavic reflex ofIE *-1:- < *-ihr . Within the context ofthe present theory, nothing would be more natural than to assume that forms such as OCS impv. 2 pI. mhnite, like dadite, continue lE prototypes in which the optative suffix was applied directly to a suffixless athematic stem. In the singular, the contrast between 2, 3 sg. mhni < *-í"s, *-í"t and 2, 3 sg. daidh < *-dj'és, *-dj'et would then reflect the fact that the ablaut *-je-/*-í"- « *-jehr / *-ihz-) was confined to the opta tive active in Indo-European, while the paradigm of the optative middle was characterized by *-í"- « *-ihr ) throughout 42

).

The Standard Lithuanian "permissive" 3 p. te-minie is in appearance a thematic optative and, like the 2sg. indicative in -í(e), is probably to be interpreted as an analogicalformation on the model ofthe ja-presents (cf. § 93). Dialectal Lithuanian, however, has permissive forms of the type te­miniJ (cf. Stang, op. cit., 422ff.) ; these, despite intonational difficulties, can perhaps be equated with their Slavic counterparts in -i. A direct comparison of Lith. te-miniJ with OCS mhni would entail the assumption that BS *-t(t), which would regularly have shortened to *-i in Lithuanian, was analogically lengthened under the influence of the *-í"- which still

41) Under this interpretation it is attractiye to compare BS *sedint- (Lith. sé'dint-, OCS sédl?st-) with the isolated Vedic participle siidat- (sadádyoni- RV 5. 43. 12).

42) Here and elsewhere, l have tacitly assumed that daidh, vizdh, etc. have been shortened from earlier *daidi, *viidi, etc., via a process typologically comparable to the shortening observable in Lat. die, düc, jac. The phonologically regular treatment of the 2 sg. athematic optative appears in OCS 2 sg. xosti = xosteli (cf. Vaillant, op. cit., 401).

116

í characterized the optative plural and dual. The new long vowel would then automaticaUy have received circumflex intonation43

).

§ 99. The foregoing discussion has shown that in all essential respects the Baltic stative type in *-i- and the corresponding Slavic type in *+ can be exactly compared with their functional analogues in the Germanic third weak class. As in Germanic, the nucleus of the Balto-Slavic category appears to have been a smaU group of lE 01' early post-lE perfect middles (*mini-, *budi-; cf. §§ 92-5); in one instance, as, m'utatis mutandis, in Germanic also, an ~-present seems to have been crcated from a middle root aorist (*siki-> SI. *shéi-; cf. § 96). This inventory has been enlarged by the secondary transfer to the ~-class of earlier je/o-presents (cf. OCS stoi- etc.; § 97) and athematic active formations of various types (*dosi-, *gari-, *y,í"di-, etc.; cf. §§ 89-90); in the latter cases the elaboration of fuUy developed paradigms in *-~- must be regarded, at least in part, as a development of the post-Balto-Slavic periodo

The most striking difference between the primary statives of Balto­Slavic and Germanic lies in their behavior outside the present system: while the preterite of Go. munan, -ai]J is evidently based on an inherited root aorist 01' pluperfect (cf. §78), the extra-present forms of *mini- are built from a stem containing the stative suffix *-e-. This morpheme has played only a peripheral role in the preceding chapters, where we have denied it a position ofimportance in the history ofthe third present class in Tocharian, the third weak class in Germanic and the presen t of the e/~-type in Balto-Slavic. N o account of any group of "e-verbs", howeyer, can be complete without a study ofthe function and distribution ofIE *-e- itself. Our concluding remarks are largely devoted to an examination of this suffix.

43) Similarly, the circumflex intonation ofLith. -ie- must be secondary yis-a-vis the acute of SI. -é-, which, as Hollifield points out to me, continues a sequence *-oihr . The retention of *-i as a long vowel in the third person would obyiously haye been favored by the fact that this ending was synchronically analyzable as stem-yowel *-i- + mood sign *-'i-.

For a general discussion of the lE optative in Lithuanian, which still presents many unsolved problems, see Stang, Vergl. Gr. d. balto 8pr, , 422ff.

117

Page 59: Jasanoff.1978.Stative and Middle in Indo-European

v

CONCLUSION: THE ROLE OF *-e- IN INDO-EUROPEAN

§ 100. Before proceeding further, it will be useful to examine our results thus far from a slightly wider perspective. It has been shown that the Tocharian, Germanic and Balto-Slavic presents traditionally regarded as primary (deverbative) "e-verbs" in fact belonghistorically to two main types. The first consists of deponent presents derived from root aorist in­jlillctives by the substitution of primary for secondary middle endings. Such forms, characteristically inchoative 01' durative rather than stative in meaning, are especially well-represented in Tocharian, where, as shown·in ch. 2, they constitute one ofthe two main so urce s ofthe attested classIII (IV) presen ts 1 ). Isola ted exam pIes of the same forma tion ou tside Tocharian are Gmc. *fulgan, *-aip 'follow' (cf. § 66) and SI. *shcati, *shéit'h 'mingere' (§ 96); the present of the latter verb invites comparison with CToch. *sikiíta1' 'overflows' and thus, as already noted, suggests the possibility that a 3 sg. preso *sik6r may already have existed in lE times.

To these may perhaps be added the lE prototype ofGmc. *dugaip 'helps, taugt' and Ved. duhé 'gives milk'. For clarity ofpresentation lE *dhugh61' . was treated in §§ 64fT. as an ordinary root present; in several respects, however, its status is exceptionaI. In the first place, the root *dheugh­underlies a thematic aorist in Gle ~·ruxov; given the close relationship between the thematic aorist and the middle of the root aorist in other ver bs (cf. §47) we should rather have expected that a form like 3sg. *dhughó would belong to the aorist, and not the present system. In Germanic, *dugan, *-aip is the unique example ofthird class weak verb which can be equated with a suffixless oxytone present elsewhere in the family; synchronically it is closely connected with a preterito-present (*daug) and an archaic weak preterite (3 sg. *duhtai; cf. § 78), both of which are normally associated with aoristic roots. Furthermore, note that although middle ~oot presents in Indo-European regularly show zero-grade when they are opposed to athematic actives with full-grade (cf. Ved. 3 sg. b1'uve, b1'iUé beside braviti 'speaks', hale (SA 12.27) beside hánti 'slays', etc.), full­grade middles are the rule in presents media tantum (cf. Ved. cá~te 'sees',

1) The other large group, of comse, eonsists of old thematic deponents in *-ske(o-, which have not played a signifieant role in the elaboration of e-statives elsewhere.

118

T !

1

váste 'wears', sáye 'lies'). lE *dhughÓ1', if a present ofthe ordinary type and not a transformed aorist, would clearly violate this pattern, since the active present represented by Skt. 3 sg. dógdhi 'milks' is almost certainly a back­formation from the paradigm of duhé (cf. §64)2).

It seems likely, then, that aorist presents of the Tocharían type already constituted an incipient class in Indo-European, where, as suggestedin § 36, they were probably created on the basis of modal, injunctive and part~cil~ial forms which were felt to be aspectually ambiguous. (This amblgmty was d?ubtless furthered, in the case of *seik- and *dheugh-, by the fact that nelther root had a contrasting present stem with a well­developed middle paradigm.)3) Similar pro ces ses continued to operate in the post-lE periodo Closely allied with the Tocharian class III (IV) presents are the sixth class presents oflndo-Iranian: these, however, seem not to be based directly on middle injunctives in *-6, but on the corresponding (thematic) active type in *-ét (cf. Av. hiéaiti, Ved. sicalp beside CToch. *sikiíta1' (§33); Ved. sp'{sát beside Gmc. *fulgaip (§66); see also ch. 2 note 29)4). '

§ 101. The second major source of "e-verbs" in the languages studied aboye is the perfect middle. We have seen in §§ 68-9 that perfect middles were already formed from the roots *men- and *y,eid- in Indo-European; if the res.nlts of chs. 3 and 4 can be upheld, anumber of other roots may belong here as well. Semantically, the perfect middle is distinguished from other middle formations by its consisten tly stative meaning; formally, it differs from the type *dhugh61' , *sik61', etc. only in the fact that it shows reduplication in Indo-Iranian, Greek and Celtic. In Germanic and Balto­Slavic, where reduplication has generally been eliminated from the perfect system, the contrast between perfect middles and middle aorist presents has beco:rne blurred. 013 sg. dugi1', in its meaning 'helps', can straightfor­wardly be compared with Ved. duhé; at the same time, however, it also functions as the North Germanic counterpart of Go. daug and OHG toug 'taugt' , and in this value is indistinguishable from the expected reflex of an lE perfect middle *(dhe)dhugh6r. Similarly, OHG gi-wizzet 'is capable' can be directly equated with Ved. vidé and 011'. 1'0 'fiti1' 'knows' while Go. witaip 'observes' has rather the non-stative sense of an aorist ~resent (cf.

2) Another possible lE aorist present of this kind is suggested by the equation ~oeh. B orotiir = Hitt. artari (§ 33); given the frequency with which preterites give r18e to new presents in Hittite, however, it is just as simple, and thus probably preferable, to suppose that m·tari is a back-formation from pret. artat(i) (= Ved. arta).

3) The nasal present represented by Ved. siñcati, if old, was presumably originally confined to the active. , 4) ,Si~ila~ly, the zero-grade thematic presents of Baltie (ef. Lith. süpa 'roeks', l~pa clImbs, etc.) appear to have developed from forms in 3 sg. *-ét.

119

Page 60: Jasanoff.1978.Stative and Middle in Indo-European

§ 69). Such semantic variation merely underscores the fact that in the "Northern" languages the types *dhughór and *(me)mnár have coalesced into a single present class, characterized by both durative-inchoative and stative functions. It is thus entirely possible that in individual cases a Germanic 01' Balto-Slavic stative may be based directly on an aorist present rather than on a true perfect middle; in effect, it is neither useful nor feasible to insist on the distinction between the two categories in the post-IE history of these languages.

Indeed, it may even be questioned whether forms like *dhughór and *(me)mnár belonged to fully separate formations in Indo-European. An aorist present of the type *mnór (01' *m~nár) could have acquired stative value through normal semantic change, just as the stative sense of lE *m~-jéjó- (cf. Gk. (J-dvo(J-aL, etc.) probably evolved from the intransitive, but non-stative meaning of jejo-presents like *mr-jé¡ó- (cf. Lat. morior, etc.). Following such a development, *m(~)nár could naturally have been reinterpreted as a perfect and provided with optional reduplication. It would then be possible to regard the perfect middle and presents like *dh1¿ghór and *sikó1' as representatives of a single pre-IE category, characterized formally by zero-grade root-vocalism and a 3 sg. in *-01', and covering essentially the same range of semantic functions as the present type in *_jéjó_ 5

). But this is speculation.

§ 102. The formations just discussed are exclusively deverbative; the same cannot be said fOl' the lE types in *-e- and *-ejejo- (i. e., *-ehr , *-ehr jejo-) , which, as we have seen in §6, supply denominative statives and inchoatives in Anatolian, Greek, Italic, Celtic and Balto-Slavic, and probably once existed in Germanic as well. This fact is alone sufficient to cast grave doubt on the supposition that *-e- was excluded fmm the present system in Indo-European (cf. § 8). A far likelier assumption is that *-e-, like other lE suffixes with denominative functions, originally occurred with both primary and secondary endings, and that in Greek and Balto-Slavic, the two branches of the family which employ *-e- as an aorist marker (cf. Gk. E(J-!X.V'l), OCS mEne), the imperfect type in *-em, *-es, *-a was su bsequent­ly utilized to provide new aorists to stative and intransitive presents of varying origins. In a substantial number of cases it can be shown that an e­aorist has replaced an earlier middle root aorist: in Greek compare such pairs as (J-[y-t¡ vs. (J-[X'rG, E7t&.y·1J vs. 7t~XTO (cf. Chantraine, Gr. hom., 400) and M'p) beside Toch. B wokota1' (§ 34); in Balto-Slavic compare OCS leza ( < *-e)

5) A similar "stative" category is postulated for Indo-European, though without attention to the specific evidence given aboye, by N. Oettinger, MSS 34, 109-50 (1976). Oettinger's discussion of the Hittite ending -a(l'i) is highly relevant to the present study, although his view of the lE verbal system differs widely from that adopted here.

120

.-.-'---------------------------------------------------------~~~~----------,

beside Gk. AéXTO, OCS b'hde, Lith. bude'(jo) beside Ved. áhudhran and Lith. gule'(jo) beside CToch. *kuliíJar (§95)6).

The morpheme *-e- is peculiar, however, in several respects. One of its most notable idiosyncrasies is the fact that, unlike other lE verbal stem­formatives, it not only functions as the basis of a regular finite conjugation, but also serves in isolation to derive a nominal form with a variety of periphrastic uses. Thus, in Latin the imperfect and future ofthe denomina­tive stative rubere are rubebam and ruMbo, respectively, which are most simply analyzed as an "infinitive" *rubefollowed by inflected forms ofthe root *bheuhz- 'be, become'. Primary statives in Latin form their imperfect and future in the same way (cf. tacZbam, tacebO); the imperfect type in -wam, more~)Ver,. recurs in the third conjugation (cf. dücebam, etc.), where -e- synchronically replaces the thematic vowel of the present stem. This situation has often been compared with that in Slavic, where the imperfect ofverbs in -eti is made by adding what appears to be a preterite ofthe root *(hj)es- 'be' to the infinitive stem (cf. OCS 1 sg. stareax'h, mEneaX'h, 2, 3 sg. starease, mEnease, etc.)7), and where thematic presents, as in Latin, show a special stem-form in -e- < *-e- before the imperfect endings (cf. OCS vedeax'h beside 1 sg. preso vedq). It isnot impossible that theproblematic Old Irish f-future (cf. 3 sg. do' moinfethat') rests on a similar formation 8).

The -e- of Lat. l'ubebO, -bam is inseparable, moreover, from the second vowel of factitive ver bs such as calefacio (pass. -fiO), arefacio, patefacio, etc., the -e- ofwhich is due to iambic shortening 9

). That such forms represent a comparatively late fusion of original two-word sequences is clear from phonological considerations alone; an inherited compound *calefaciowould have been weakened in Archaic Latin to *caleficio, like pe1'-, con-ficiO, etc. Note also the following ear ly examples of factitives with tmesis : ferue bene facito (Cato, 1'. 1'. 157, 9), peljeme ita fiet (Varro, r. r. 1, 9, 2), conS1¿e quoque faciunt (ibid., lI, 9,13), excande mefecer'unt (ibid., lII, 4, 1),facit are (Lucr. VI, 962). Such cases directly refute the attempt by Skutsch, Kl. Schr. 214, 285, to derive calefio from *calensfio, and to explain calefacio as a back­formation from calefio. Likewise unattractive, although for other reasons,

6) Note further that under the analysis set forth in § 78, the Balto-Slavic aorist in *-e- (3 sg. *minet, etc.) maypartly correspond in Germanic to a reflex ofthe lE root aorist (3 sg. *mundai).

7) For theories of the Slavic imperfect, the second element of which is obscure, see, e.g., Meillet, Sl. com., 272ff. and Vaillant, op. cit., 66f.

8) Numerous explanations, of course, have been put forth to explain the lrish f-future; the most attractive, in my view, is still that which derives 'moinfetha1' ~rom *man"í-fJha- where *man"í = BS *mine-, Gk. fW.V·~-, and where *fJMi- is the lrregularly syncopated reflex of *besa- (for *esií-, a-subjlIDctive of the copula; cf. Lat. erat). See Thurneysen, op. cit., 398.

9) In iírefacio, of course, we fmd the generalization of -e- from calefacio, patefacw, liquefacwand other verbs with a short initial syIlable. The etymological-e- is still found, e. g., in consuejacio (Ter., Ad. 1.1. 29, etc.).

121

Page 61: Jasanoff.1978.Stative and Middle in Indo-European

is Leumann's explanation of calefit as a replacement of calescit after the latter had been folk-etymologized as cale + escit (1F42, 62 [1924]): escit simply does not have the required meaning 'becomes, is becoming' in Old Latin.

Latin thus utilizes denominative stems of the form * X-e- in three ways: 1) to make stative presents meaning 'be X' (rube1'e, calere) , 2) to make periphrastic imperfects and futures with the root *bheuh2-, meaning 'wasjwill be X' (rubebam, -bO, calebam, -bOJ, and 3) to make factitives with the root *dhehr , meaning 'make X' (calefaciO). As we shall now see, this configuration of functions has a striking parallel in Vedic Sanskrit.

§ 103. The Vedic root guh- 'hide' (pres. gflhati, -te) is associated with an adverb gúha 'hidden, in hiding'. This form is attested a total of 53 times in the Rigveda, 22 of them in conjunction with the roots dhéi- (ni-dhéi-) and kr-. The sense of gúha (ni- )dhéi- and gúha lcr- is 'make hidden', as the following examples show:

1) 1. 130. 3 ávindad divó níhita'f(b gúha nidhí'f(b vér ná gárbham páriv'itam áá1nany anánte antár áá1nani

'Er fand den im Versteck verborgenen Schatz des Himmels, der im Fels verschlossen war wie die Brut des Vogels (im Ei), im endIosen Fels' (Geldner).

2) 2. 12. 4 yénemá víáva cyávana lcrtár¡,i yó dása'f(b várr¡,am ádhara'f(b gúhálca'"

'Durch den (1ndra) alle diese UmwiUzungen geschehen sind, der die dasische Rasse unterworfen und verdunkelt hat .. .'

3) 5. 15. 5 padá'f(b ná tayúr gúha dádhano mahó rayé citáyann átrim aspa'"

4) 10. 5. 2

'Del' du wie ein Dieb deine Spur verbirgst, du hast jetzt zu groBem Reichtum dich offenbarend dem Atri (aus der Not) herausgeholfen. '

rtásya padá'f(b lcaváyo ní panti gúha námani dadhire párar¡,i

'Die Seher hüten der Wahrheit Spur; sie haben ihre h6ch­sten Bezeichnungen in ein Geheimnis gehüllt.'

Used independently of (ni)dhéi- and kr-, gúha characteristically has the value of a predicate adjective, with qr without an explicit copula:

122

5) 3. 1. 9 gúha cámnta'f(b sákhibhi'" áivébhir divó yahvíbhir ná gúha babhüva

'1hn, del' VOl' seinen guten Freunden sich verborgen hielt­vor den jüngsten T6chtern des Himmels war el' nicht verborgen.'

6) 8. 14. 8 úd gtt ajad ángirobhya aví~ krr¡,ván gúha satí'"

'El' trieb die Kühe den Angiras' heraus, die versteckten zum Vorschein bringend.'

7) 9. 102. 2 úpa tritásya paSíor ábhalcta yád gúha padám

'In des Trita Kinnladen (?) hat er seine geheime Stufe erreicht.'

8) 10. 45. 2 vidmtt te niÍma paramár¡~ gúha yád vidmfÍ tám útsam yáta ajagántha

'Wir kennen deinen h6chsten N amen, del' geheim ist; wir kennen den Quell, von wannen du gekommen bist.'

The syntactic overlap of gúha in these examples with the Latin "infinitive" in -e is too remarkable to be ignored. The locution gúha dha- (lcr-) resembles the Latin type calefacia, facit are, etc.; the use of gúha with forms of as- and bhü- recalls Latin imperfects and futures like calebam and caleba. Only the Latin conjugated type in -ea, -es, -et, etc., lacks a precise equivalent in Vedic: its place is taken by the predicative use of gúha without a copula as in exx. 8) and 9), where the phrase yád gúha (gúha yád) is translationally equivalent to Lat. quod latet « latere).

The comparison of gúhawith e:-forms elsewhere takes on added interest when it is noted that the root *gheu[jh- actually underlies an e-stative in BaItic. Ved. gflhati (cf. also YAyo 1 sg. mido aguze '1 hid (myself)', etc.) is cognate with Lith. guzti 'cover (with something warm)'; this verb in turn underlies a stative güzeti (3 p. guza, guzi, güieja) 'líe (under something warm)', typically used of young birds nestling beneath their mother. Clearly, it would be desirable to relate gúhato the stative stem *ghu[jh-¿- in some direct way.

§ 104. Unlike Lat. *cale, are, etc., the morphological structure of gúha is clear: it is simply the instrumental singular in -a « *-e < *-eh1 ), with adverbial accentuation, ofthe root noun guh- 'hiding place, concealment'

123

Page 62: Jasanoff.1978.Stative and Middle in Indo-European

---~ ~"------------------------------------------

(cf. acc. sg. gúham RV 1. 67. 6). The literal meaning of gúha is thus approximately 'with (i. e., in) concealment'; for the instrumental in this value compare, e. g., the Latin "ablative of description", originally probably an instrumental also (cf. Delbrück, Vergl. Syn.240). From a typologicalpoint ofview, nothing, sofar as 1 am aware, standsin thewayof identifying the *-e of such forms with the -e- (> -13-) of Lat. calefaciO, calebam and calebo, and the -e- of OCS mbneaX'b. Once established as a regular constituent of the verbal paradigm, of course, the instrumental "infinitive" would have tended to become derivationally attached to the present stem with which it was associated; it is neither necessary nor desirable to assume an instrumental *deuk-eto account for the genesis of a form like Lat. dilcebam 10).

Perhaps even more significant is fue fact that instrumentals like gúha can be used to explain the creation in Indo-European of the fully conjugated stative type in *-emi, *-esi, *-eli, etc. We have alreadynoted the collocation yád gúha (gúha yád) in exx. 7) and 8) aboye, where gúha is the syntactic equivalent of a stative verb; particularly striking is the parallelism of yád gúha in the first pada of ex. 8) with yáta ajagántha in the second. Similar instances are provided by the Vedic adverb mríJa 'in vain' (cf. Hitt. madezzi 'is false, deceptive' < *m'[s-e-) , which is used predicative­ly both in its only Rigvedic occurrence (1.179.3 námríJasrantá1f1, yád ávanti deválp 'nicht vergeblich ist das Mühen, das die G6tter begünstigen') and later (e. g., AV 5.18.9 yám ásyanti saravía1f1, ná sá mríJa 'the arrow which they send isnot in vain'). It may be conjectured that in such constructions certain late lE instrumentals were reinterpreted as unmarked 3 sg. verb forms; they would thus have become capable of receiving personal endings when employed with non-3 sg. subjects. A new 1 sg. in *-emi and 2 sg. in *-esi created in this way would have stood in the same relationship to the 3 sg. in *-e as forms like PoI. pret. 1 sg. znalem '1 knew', 2 sg. znalcsz stand in relation to the endingless 3 sg. (historically a participle) znal ll

). Subse­quently, *-e would have been replaced by *-eli in the 3 sg. under the influence of other athematic paradigms. Probably an important factor in the reanalysis of the instrumental ending *-e as a verbal suffix was the growing obsolescence in late Indo-European of the root noun class from which forms like gúhawere derived, and its replacement by the productive neuter type in *-e/os- (cf. Gk. yévo~, Lat. genus < *f¡énhre/os-, Ved. tápas-

10) Nor, of course, need we assume an actual case form *1}edh-e(*-eh1) to explain OCS vedéax'b. The Latin first conjugation imperfect in -abam, like the Slavic imperfect in -aax'b, is perhaps best explained by a proportion of the type tacere (mbnéti) : tacwam (mbnéax'b) :: cantare (délati) : x, x = cantiibam (délaax'b).

11) Or, conceivably, a 3 sg. pret. in *-€l could have been created to the present 3 sg. in *-e(with loss of*-h1 alreadyin late Indo-European) on the model ofpairs like 3sg. *bhere (pres.) vs. *bheret (impf.). The typological importance of the Polish preterite paradigm has been stressed by Watkins, Gelt. Vb., 94.

124

y'--------------------------

!

'heat' < *tep-e/os-, etc.). In the case ofthe root *gheugh- in particular, the corresponding adverb in *-e was "verbalized" in Baltic (cf. güiéli) but not in Indic, where the survival of the root noun guh- provided a syn­chronic support for the continued analysis of gúha as a nominal form 12).

While admittedly speculative, the foregoing account directly motivates two notable peculiarities of the presents and aorists in *-e- of the attested languages, viz., their failure to display paradigmatic apophony (cf. § 56), and their almost exclusive confinement, despite a predominantly medial (and, in Greek, even passive) function, to the active voice. (Impersonal passives like Umbr. loufir 'lubet' and Lat. calelur constitute arare exception.) Note also the fact, drawn to my attention by Schindler (personal cOPlmunication), that a derivation of *-e- from the instrumental singular of a verbal abstract would accord well with the role played by denominative statives in archaic "Caland's Law" alternations such as that seen in Lat. rube:re, 011'. . ruidi, etc. « *h1ntdh-i (*-éh1)): Gk. epuep6~, Lat. ruber, etc. « *h1t'1tdh-ró-) : Ved. ntdhi-kTá (PN) 'blutausstreuend' « *h1ntdh-i-; cf. Wackernagel-Debrunner Altind. Gr. n.1. 61), etc. 13). As Schindler will demonstrate elsewhere, the apparent substitution of suffixes in such forms is best explained by assuming each term in the Caland series to have been derived independently from an underlying root noun.

§ 105. If the aboye theory is correct, we may summarize the ulterior history ofthe "e-verbs" as follows. Late Indo-European possessed a series of presents and imperfects in *-e-, based historically on the instrumental singular of root nouns and hence ultimately denominative. In cases where the underlying root was not also utilized to derive a primary -verb, this type (01' íts thematized equivalent in *-e-je/o-) was able to maintain its identity in a majority ofthe lE daughter languages, giving rise to the denominative statives and inchoatives discussed in § 6. In cases where there existed a competing verbal paradigm, on the other hand, the fortunes ofthe type in *-e- were mixed. In Balto-Slavic the imperfect in *-e- was retained as a preterite ("aorist"), but the corresponding present class was replaced by the productive formation in *-r-, an outgrowth of the lE perfect middle; similarly, in Greek, the e-imperfect was utilized as an aorist, while presents such as fW.lvo[J.cl'.L, XiXlpw, etc. brought about the general elimination of forms of the type *[J.áv'l)[J.~, *Xáp·r¡[J.~ from the present system 14). This process was

12) 1 am indebted to Schindler for this explanation of the survival of gúha. 13) The commutability of stative *-e- with *-1'0- and other Caland's Law suffixes

was first noted by Watkins, TPS 1971, 64f.; additional examples are discussed in the important 1976 Harvard University dissertation of Alan Nussbaum Galand's "Law" and the Galand System. '

14) An exceptional case in which Greek appears to retain an e-present beside a non-denominative perfect is {nyéw « *-ejo, like 8IXp~éw, av8é:w; cf. § 6) beside pf. ~ppLylX.

125

Page 63: Jasanoff.1978.Stative and Middle in Indo-European

carried further in Anatolian, where *-e- was generally restricted to synchronically denominative functions, as well as in Indo-Iranian, Ger­manic and Tocharian, where *-e- was not only unable to maintain itself against the perfect middle and root aorist, but was eventually replaced in the denominative class as well. Only in ltalic was the synchronically de­verbative present type in *-e- retained on a large scale.

126

T

INDEX OF FORMS CITED

(Numbers refer to paragraphs; superscript numbers refer to footnotes.)

abhaksi 32 ábhakta 32 ábudhmn, -mm 9, 95, 102 ádhat 5410

áduha(t) 47, 62, 6431

áduhi 4444

(á)gan 32 (á)jo~i 38 áju~ran 3, 38 alipat 34 atipta 34 ámata 3, 9, 47, 68, 72 amitmyáti 83 (á)mucat 33 ániti 25 aramahi 3528

amnta 3528

amta 3528

á1·ta, arta 33, 3528 , 47, 1002

asa- 27 Mat 54 ásati 47 ásayat 47, 62 asican 33 ásthimn 3 asvñyáti 6 ásyati, -yant- 88 (á)va1·ti 38 avidanta 3528

(á)vidat 33, 47 ávismn 33 (á)vrtmn 38 áyate 43 bhárate 46 bhámti 47 bharethe 46 bharádvaja- 76 bhavi~yánt- 88

Sanskrit

bhra(rIJ)s-, bMarIJsati 33 bhramsa- 3327

bhmiat 33, 35 bhrjjáti 37 30

bibháya 2, 62, 90 bibhéti 62 brav'iti 100 brháduksan- 76 b1'1tve, b~üté 47, 100 búbodhati 71, 95 bubudhaná- 71, 95 budhaná- 9, 71, 72 budhánta 3528 , 95 búdhyate 9, 72 cakára 2 cáo$(e 1, 47, 100 dadhársa 2 dadháu' 90 dadrhaná- 95 dáhati 30 dasyámi 87, 88 dasyánt- 87, 88 diJlívara- 76 70

dhamyátkavi- 76 dhrsádvarna- 76 dhr~jj'rfá- i 5 dhro$ánt- 75 dógdhi, ádhok 64, 100 drhyati 95 duhaná- (dú-) 64, 75 duhánt- 64, 65, 75 duhánti 64, 65 duhé 47, 64, 65, 66, 100, 101 duhré, -até 64, 65 éti 73 gácchati 32 gaméma, -mahi 35 28

grdhya- 9

127

Page 64: Jasanoff.1978.Stative and Middle in Indo-European

g'f1Játi 73 g'f1Je, g'f1J'ité 73 guh-, gúham 104 gúha 103, 104 g1ihati, -te 103 hánti 100 hate 100 inóti 73 Tse 74 62

jagám 2 jaghána 2 jajana 4 jajña1¿ 73 jajñé 4 janami 30 jásamana- 30 jasyata 30 jáyate 30 jujó<$a 71 jújo<$ati 3, 71 j1ljU<$?irJ,á- 71 juju<$é 71 j1/{ju<$úl¡,71 j1ljuIIVán 71 j11<$a1Já- 71 kaviyáte 86 k<$éti 1 k1lpyati 97 leláya( ti) 62 limpati 34 lúbhyati 9, 72 mamársa 3 mamn¿t(h)e 68 mananá- 9 mányate 9, 12, 53 8, 68, 72, 83, 93 32

rnriyáte 82 mflla 104 m'f<$ánta 33, 35 28

mr<$thál¡,33 mli<$yate 3 m'fta 96 m1lñcáti 33 m1l<$1Játi 31 námate 30 nasanti 30 náyati, -te 36, 43 nudáti, -te 32 mdthal¡,32 pacati 30 pácyate 30

128

paprátha 4 papmthé 4 patsi 32 patthal¡,32 pávate 73 p'f<$ánt- 39 punáti 73 pÚ<$yati, -yant- 53 8, 88 riÍsva 72 r'ikthál¡,3 rócate 3, 33 'ftasp¡is- 66 ruciiná- 33 ruciya 33 rudhi-krá- 104 ruróca 3 rusátpas1l- 76 sadádyoni- 97 41

siÍkllVa 72 sána 83 8

sanayá- 83 8

sankate 43, 72 sásayana- 71 sáyana- 71 Sáye 1, 47, 71,100 sécate 33, 96 sical¡, 33, 100 stdati 97 si1icati 1003

si<$iÍya 54 Sóbhate 36 sóbhe 36 sprsát(i) 66, 100 sraváyati 39 sr1JÓti 39 Sróljamana- 33 Stá1di 38 stávate 72 stáve 38, 72 stuvánti 38 sucádratha- 76 SUlj1lpaná- 96 S1lljUpthal¡, 96 svapiti 96 sváptu 96 Svitaná- 95 syáti 97 tákljati 38 tápas- 104 tastháu 3, 85

tiÍlj(i 38 tatrpaná- 95 tat¡;pé 70 tatrlja1Já (ta-) 71 tatr<$úl¡,71 tmsádasY1l- 76 70

t¡;pnóti 38 trlj?irJ,á- 9, 71, 72 tfljyati 1, 9, 72 uk<$anyá- 1 uparispts- 66 váidadaSvi- 76 vámsva 72 va~áti 72 ' vanéma 72 vanta 72 vár'tate 3 váste 100 v~ti 1 vavárdha 4 vavárta 3 vavrdhaná- 75

aguze 103 ahiSaiia 2 bamt, zaoOra- 76 büioiia- 9 büsiiant- 88 calte 1 caxnar'¡¡ 2 da!aiti 30 d'iJr'iJzano, p'iJr'iJOa- 95 dm!aite 95 fmsicanti 33 hazdiiat 97 38

hicaiti 100 hioaiti 97 iriOiieiti 32 mamanaite 68 mamnanai 68 mamne 68, 71

akamas 21 ar-, aratar 18, 33 arlilj 41,44 ar~ 33

véda 2,3,69 vidádvasu- 76 vidán 35 28

vidaná- (ví-) 69, 75, 76 vidánta 35 28

vidata 35 28

vidé 69, 71, 97, 101 vidh-, vídhyati 32 vidmá 3 vidré 69 vidviÍms- 69 46

vijáte '33, 35 vikta 33 vindáti 33, 47 55

visáti 33 vitse 69 vrdhaná- 75 vrdhánt- 75 yátati 38 yodhaná- 32 yúdhyati, -te 32

Avestan"

manta 3, 9, 47, 68 n'iJmaiti 30 n'iJmante 30 paéaiti 30 raocaiti 33 srauuaiieiti 39 tasti 38 tütauua 2 vaunus 72 vidat, gau1¿- 76 vioat, xVar'iJnah- 76 vinasti 47 55

visaite 33, 36 29

yaetul 38 yaiiata 38 yateiti 38 yüioiia- 32

Tooharian A

araljt 41, 44 arsam 41 art- 18 as-, asatar 17,18,19,27,37

129

Page 65: Jasanoff.1978.Stative and Middle in Indo-European

asa,! 27 ciñca1', ciiiicar 23 cki'icar 23 cmatar 16, 30 cmol26 epe 25 18, 26 19

kan- 16, 30 karema11J 15, 39 kare,! 3937

kary- 15, 39 karya,!39 karyeñc 39 kas- 16, 30 !catiiñki'i,! 32 kla(w)- 18 !claw- 18, 39 klawaéf 39, !clyosantar 20, 21 !clyosma11J 28 !clyosmar 28 !clyo'!tar 20 !cnatar 30 !cosne 25 kowi 26 !cul-, !culatar 15, 19, 33 k~li 33 !culyp- 15, 23 kur- 15 lac 35 la(n)t- 35 lit-, lita(n)ta1' 15, 18, 32, 38 litk- 32 lotk- 18 maca?' 20 mal- 15 mañ 19 martk- 32 miisk-, maskatar 14, 15, 21, 31, 43 miiskant 39 maskas 14 ml1¿sk-, mloskatar 18, 31 nak- 16, 30 22

natk- 32 natsw- 15 ñom 26 okat 24, 25 onk 26 onkalam 26 opyac 39 01' 24, 25

130

pacar 20 pak- 16, 30 pakna'!tar 30 23

pal-, platar 15, 37 palk- 23 palsk- 37 30

park-, parkatar 15, 19, 33, 66 piirsk-, praskatar 18, 31 pats 21 pketar 30 plant-, plantatar 17, 18 plantmar¡L 18,28 plantmar 28 poke 24 pons 25 pot-, patatar 18, 27, 38 pracar 24, 25, 26 prank 21,27 2°,33 praski 26 putk- 32 ritw- 15 ritwant 39, 65 ritweñc 39, 65 rtar 23 rwa- 20 sak- 18 satk- 15, 32 sik- 15, 33, 96 spaltk- 66 sparcwant 39 sparcw,!-a11J 39 spartw- 18 sral, sralune 16 stwar 21 SU1'- 15 wa- 20 tam- 16, 30 tiisk- 15, 18 tiis!cmam 18 tpuk- 15 trap- 18 fj'e 39 trik-, trikata1' 7, 15 trikant 39 triw-, triwatar 15, 37 triweñc 39 tsak- 16, 30 tsiilp-, salpatar 18, 20, 38 tsam-, sama(n)tar 18, 20, 38 tsankar 38 32

tsar-, tsratar 15, 16, 30 tsarama" 39 35

tsarw-, tsanvata1' 18, 38, 70, 95 tunk 26 wakaf¿ 34 wal- 16, 30 wank- 18 want 19, 20 wiip- 15 war 21 was 23 wasankatar 32 Waéft 24, 25, 26, wat 23 . watk- 15, 32

aiw-, aiwotar 18, 38 alce, akenta 22, 23 akem(o) 21 aniisk- 25 ar-, orotar 18, 33, 3936, 1002

ara 33 art- 18 as-, osotar 18, 19,27,37,63 i'isii'll$a11J 27 camel26 ciñcare, cañca1'e 23 cmeta1' 16, 30 e1ikwe 26 epe 25 18,26 19

epiyac 39 e1'l¡ar¡¿ 41,44 ersau.41 e1',!t(o) 41, 44 kalpaskentar 31 kalpal$,!a11J 43 kiilpi'istii1' 31, 43 kan- 16 !carp- 18 kii.s- 16, 30 kery- 15, 39 keriye11J 39 keriyemane 15 Jcewiim 26 klaik;- 18, 38 klapalñe 39 36

klautk- 18, 32 klaw-, lclowotar 18, 3936

Jclawa 3936

wik- 15, 33 wlatii1' 30 yam-, y(p)a- 39 35

yat-, yatatiir 18, 38 ye1'pe 25 18

yme 39 yo!c- 24,25 yoke 39 yom- 25 yow- 25 ysomo 25 Y1¿- 15, 38 yuk 23, 26 yutk-, yutkatiir 15, 32

Tocharian B lclay- 18 lclyausemane 28 Jclyausemar 28 !clyausentiir 20, 21 klyau'!tiir 20, 33 klyelle 33 klyiye 33 knetiir 30 kas 25 lcramp- 15 ksetii1' 30 kul-, kuletar 15, 19,33,95 kulyp- 15, 23 kurp- 15 kwiilypelle, kulypelle 23 kwiir-, Jcwremntiir 15 lac 35 ' la(n)t- 35 lait-, laito(n)tar 15, 18, 27, 32, 38 lip-, lipetiir 15, 19, 34, 72, 95 lit- 32 litk- 32 l1¿-, lyewetiir 15, 20, 38 luk-, lyuketiir 15, 19, 33, 66 lutassiim 32 24

lyatdcsa' 33 lyu- 15 macer 20 ma1ik- 15 mars- 15, 19, 33 ma1,tk- 32 miisk-, miisketiir 14,15,21,31,43 mas!ca 14

131

Page 66: Jasanoff.1978.Stative and Middle in Indo-European

miiskeñca 39 matsts- 15 meñe 19 mit- 15 musk-, musketiir 15, 31, 43 niik- 16, 30 22

nakiiTIJ 16 niim- 16, 30 ñiisk-, ña~tiir 187

niitk- 32 ñem 26 nesiiTIJ 2 nkem 16 nmetiir 30 ñmetsi 30 nu-, ñewetiir 15, 20, 38 olet 24,25 onkolmo 26 onolme 25, 27 0'1', arwa 22, 24, 25 ost 24, 25, 26 pacer' 20 piik- 16, 30 piilk- 15, 19, 37 piilsk- 37 30

paut-, pa1dotiir 18, 27, 38 petso 21 pUe, pilenta 22 pilko 23 pliink- 15 plant-, plontotiil' 17, 18 pliitk-, plyetkemane 15 pokai 24 poñc 25 postaññe 25 priink-, priinketiir 15, 33, 35 prerike 21, 27 20, 33 procer 24, 25, 26 proskiye 26 pl'utk- 15, 32 putk- 32 pyamttsait 46 mtr'e 23 ritt- 15 nlwa- 20 samp- 18 siitk- 15, 32 sem 32 sinastiir 30 23

80y- 20 9

132

spalk- 66 spiint- 15 spantii~liiiTIJ 32 24

spiil'k- 15 spartt- 18 spaw- 18 speltke 66 sruk-, sruketiir' 15, 37 stare 92 30

ste 3 #wer 21 wa20 syelme 25 trom- 16, 30 tanlcw 26 tas-, tasaitiir' 15, 18 tkacer 23 tmi 39 trapp- 18 trile-, tl'iketiir 7, 15 triw-, triwetii1' 15, 37 tsiik- 16, 30 tsiilp-, tsiilpetiir 15, 18, 38 tsiim-, tsmetiir 15, 18, 38 tsii'ñle-, tseriketiir 15, 3832

tsiir-, tsretiir 15, 16, 3935

tsarw- 38 tsarwoytiir 38 tsriille 16 t,m- 15 t1lle-, c1lkemar 15 wale-, wokotii1' 18,19,34,3936 ,63,66,102 wiiks- 15 wase 23 wate 23 were 21 wik-, wileetiir 15, 33, 35 yakwe, yiikwenta 23, 26 yiim-, yamastiir 187

yanle- 18 yat-, yototiir 1 8, 38 yente 19,20 yerpe 25 18

ymiye 39 yok- 24, 25 yoleiye 39 yom- 25 yop- 25 ysomo 25

'..,

alclJi 4 akkis1cattari 43, 46 appant- 39 ares1cattari 43 a'rtari 1002

artat(i) 3, 1002

asasi 62 ellJat( i) 4444

el(t)ari 62 lJaliya(ri) 36 lJalzai, -iyanzi 12, 54, 84 lJalziwen 54 J2

lJark(iya)- 73, 82 lJami(n)le- 73 lJaMuet 6 lJ,uwai- 54 ishiii, -iyanzi 54, 97 islJamai- 54 i8lcalliii(zzi) 62 ispiii- 54 iyattari 43 ganeszi 73 56

lca-an-lea-an-zi 43 41

gangaMi, -ai 43 Jcanlei (lea-a-an-ki) 3834, 43, 72 karp- 33 learpis1cattari 43 leitta(r'i) 1 leurkuris1cattati 41 liilei 95 lazziyattar'i 41 luletat 33 miii 54 mallai 3834

marsezzi 19, 104 martari 96 naMariya- 1 niii- 54

1l(wa)lantis 30

ba1:nam 33

Hittite

nakleeszi 6 nakleezi 6 newalJ- 74 neya(ri) 36 paM-6 pai- 54 pai8lcatta 43 pa~is-ga-ta-ru 43 40

pais1catt1lma 31 parai- 54 par'kelzi 19 parkl:ya- 33, 66, 82 padctaru 33 pU1'siya(zi) 62 piwen 54 12

piyant- 39 sai- 54 salelJi 2 sarmttari, -dd1lma 41 sa1'ratti, -tteni 41 suppa1'i, s1lptari 96 dai, tiyanzi 54 diiitti 54 12

dalek- 187

tannatteszi 6 tannattezz'i 6, 54 tarniii(zzi) 62 teMi 54 tezzi, tet 54 tiya-, tiyazi 41, 97 dus1cis1citta 31 ui8lcitta 31 wiilei 34 wallJ,zi 30 welezi 1 ziii- 54 zikle- 32 26

Luvian

Armenian

bemy, -ar 89

Page 67: Jasanoff.1978.Stative and Middle in Indo-European

ebarj 33 egit 47 elcn 32

af3f3E1!ETOe 46

djelc 30 kam 59

(F)áY¡'VTW 7 aOtXl)15Et 47 aeTW 20 9

aCw 27 aAoxo~ 33 av()éw, -ijafll 6, 105 14

ae07:eOl' 55 aeów 55 aexéxaxoc; 76 70

f3atvw 32 f3áAAw 33 f3 a(ftAeVW 1 f3ef3Ar¡¡IIÍVO~ 71 f3éf3A/lTfll, -TO 71, 95 f3AijTO, -PE¡'O~ 33, 71, 95 yéyol,a, -e 3, 4 yéyef1Cpa 2 yévo~ 104 ytyVO{1fll 3 oéowxa2 Ostxaet 47 oé¡wc; 38 oé¡lO) 38 oovA6w 74 (F)é(F)iíye 1, 34 e(f)áYI), (F)ayi'¡l'fll 7,19,34,63,102 e(f)coe 47 efxTo 71 elAlíAov()e 47 e'l'lfleTw, -TO 4, 71 ef¡lfle¡tÉJooc; 71 éxdJ1' 1

134

gtanem 47 55

wtim 6

Phrygian

Albanian

pjelc 30

Greek

€flávr¡, pavi'¡l'w 7, 50, 54, 63, 102 e¡c¡weE 4,71 eJ'Ve:rre 58 eotxe 71 [m;áyl) 102 env()eTo 95 e(!tcw 1 eeetya 105 14

e(!v()eóc; 104 EaTr¡xa 2 eTAií 72,95 (e)TvxO¡', -e 64, 100 hAv()e 35, 47 ()aef3éw, -i'¡aw 6, 105 14

()é(!Oflfll 85, 90 (J¡r(¡axw 68 42

(F){ope¡' 3 rcw 97 38

XE!TW 1 xevów 74 xove{Cw 1 xTt¡tel'o~ 1 Aelnw 64 AÉXTO 95, 102 AÉAome(l') 3 AeAoxvla 95 ¡lfl{¡'O¡lflt, -eTat 3, 7, 9, 68,83,93 32,101,

105 ¡té¡tr¡l'a 3 flé/lV/)Tf11 68, 71 pé¡w¡'a, -e 2, 3, 68 pbw 31

piy)) 7,34,102 ¡tlxro 34, 102 ¡u¡tVataxw 68 42

¡u¡tvlíaXOpfll 68 ¡tlayw, -STW 7, 37 30

;E(C)¡,tCw 1 (F)olOa, -E 2, 3,47 oUV¡U, -¡CW 3 oAw},a 2,3 O(!Óer¡TW 4 O(!WeE 4 OVAr¡ 30 nbrEla¡lfll 4 ninoc()a 2,' 4 ninvaTw, -ro 95 nijxTo 102 étyiw, -l}afll 6, 105 14

TaenW¡tE()a 95

aeet1tm 56 albeo, -ere 1, 10 anatis 27 are 102, 103 al'efaeio, -ere 102 al'eO, -ere 19, 27, 37, 63 aro, -are 55 audio, -is, -ire 9 axim 88 ealroam 102, 103, 104 calebO 102, 103, 104 calefacio, -fio 102, 103, 104 caleo, -ere 102 caletul' 104 cantabam 104 10

canto, -are 104 10

capio, -is, -ere 9, 94 36

capire (lt.) 94 36

capso 88 carcaris 27 conS7te 102 consuefacio, -ere 102 9

coquo, -ere 30 C7tnctor, -ar'i 72 54 cupio, -'ire 97 dic 9842

d'iuido, -ere 32 due 98 42

d()¡'I)Xa2

Telew, Ter¡TÓC; 84 denw, -Opat 38, 64 deaO¡lflt 64 T8e1ptflf3(!oro~ 76 70

T8T8VXWC; 4 drAapev 72, 95 T8rAr¡vill 72 dTvxTal4 T8ÚXW 64 TVYXávw 64 <paíveTf1t 7 <pávr¡ 7 <pieet 44, 47 <pé(!eTf11 46 <pcAÉw, -ijafll 6, 83 xa{ew, -el, -8{)' 7,39,105 Xá(1) 7

Latín

ducroam 102, 104 emo, -ere 25 erit 47 escit 102 e7tidens 39 excande 102 fac 9842

faxim, -o 88 fente 102 -ficio, -ere 102 flo, -are 37 fulgeo, -ere 19, 37 geno, -ere 30 genor, -'i 30 gemts 104 (g)nau'i, -it 2, 73 habeo, -ere 7, 8, 54, 74 habit7tS 56 iaceo, -ere 33 iacio, -el'e 33 inseq7te 58 inuideo, -ere 90 iubeo, -ere 32 latere, -et 103 liquefacio, -el'e 102 9

lubere, -et 9, 63, 72 luceo, -er'e 19 maneo, -ere 7, 8, 31

135

Page 68: Jasanoff.1978.Stative and Middle in Indo-European

--~---~ ------~--~ -

mellit1ls 86 16

memini, -it 2, 3, 68 molo, -ere 38 monitus 56 morior, -í 82, 101 m01teO, -e1'e 31 nanc'iscor, -i 31 nigreo, -ere 6 nigresco, -ere 6 obli1líscor, -i 31 odi 2 oletum 56 paciscor, -i 31 pasco, -ere 6 patejacio, -ere 102 perjerue 102 placit1ls 56 postrem1lS 25 puto, -are 32 rubebam 102 rubebO 102

censazet 88 didest 88 didet (Vest,) 88 jeret (Marr,) 88 jiiet 88 20

just, j1/rent 88 loujir 104

ad'ágathar 2 'ailegedar 1 airid 55 'ber(a)r 36 rofetar 69 rofinnadar 69 45

rofitir 69, 71, 101 'gaibitl, 'gaib 9 'gainethar 30

136

rubeo, -e1'e 6, 7, 50,83,102,104 1'uber 104 1"1lbeta 56 1'UO, -ere 20 sed'i 97 seneo, -ere 6, 7, 50, 83 senesco, -e1'e 6 sileo, -ere 6, 50, 74 silesco, -ere 6 tacebam 102, 104 10

tacebo 102 taceo, -ere 7, 50, 63,74, 83,104 10

tacit1ls 56 tetul'i 72, 95 torreo, -ere 9 uehens 39 1telim, -is 90 uenio, -í1'e 32 uideo, -ere 50 uoueo, -ere 56 uoln7ls 30

Osean

just 88 sent 88 20

staiet 88 20

tlincter 46

Umbrian

tavez, tases 56, 74 vujetes 56

Old Irish

gigis 88 ic(c)- 84 10

laigid, 'legat 84 10

'léici1l, 'léici 9 11dd 35 melid 38 memais 88 do'ménair 68, 69 45 , 71 do'moinjethar 102

do'moini1l1', -ethar 9, 46, 68, 69 45

rái'd 84 re?:ss 88 'ruidi 6, 104

gwyr, gwybot 69 44

ga-ainan, -aip 74 arjan 55,84 baim 41 , 4444

bairada, -aza, -anda 29, 41 bairis, -ip 41 bauaida 77 71

bauip 77 71

ga-dars 2 daug, dugan 64,67, 101 dauhtar 55 jaheps 75,52,56 jastan, -aip 50, 74 jaurhts 31 gahugds 56 haba, -am, -and 7, 11, 52, 55 habada 50 habai (impv,) 75 habaida, habaips 51, 77 habais, -aip 51, 52 ha,ban 49, 50, 51 habands 52 haba1l, -ai(s) 75 hahip 3834, 72 hahan, -aip 50, 72 hausja, -eip 9 kann 67 keinan, -noda 73 ga-k1tnnan, -aip 67, 73 uj-k1tnnan, -aip 73, 77 uj-k1tnpa, -kunnaida 77 lais 73 ga-leipan 32, 38 liban, -aip 19, 34, 72 11lbains 9, 72 mag 67 malan 38

saidid, sedait 84 10

saigid 84 10

seiss 88 'úraigedar 1

Middle Welsh

kymyscetor 46

Gothie

man 2, 3, 67, 68 maurnan 73 munan, -aip 50, 67, 99 munands 98 nasjip 55 ogan 67 reimn, -aip 62 salbond 52 saurga 74 saurgan 74 setum 97 ana-silan, -aip 50, 74 sweran, -aip 50, 74 pahan, -aip 50, 74 pahta 52 7

parba 70 ga-pal'ban sik 67 pa1'bs 70 parj 38, 67, 95 pa1lrsjan sik 9 piwan, -aip 1, 50 ana-piwan 74 p¡tlan, -aip 50, 72 unagands 67 unwunands 50,72 wait 2,3,67

'waka 51 wakan 50 ga-waknan, -nip, -nada 73 weihan, -aip 50 weitwops 69 46

wiljau, -ei.., 90 winds 52 wit(tn, -aip 50,67,69, 101 wit1lm 3

..

137

Page 69: Jasanoff.1978.Stative and Middle in Indo-European

Old NOI'se habela, gihabel 7, 77 momen 73 (Old Icelandic unmarked) habu 49,59 roten, -el 6, 7, 50

hangiTn 50, 72 mtela 7 beffwa (OSw.) 62 olla 77 72 hapta 49, 60, 77 sagen 49 bifa 50, 62 róa 84 hebis(t), -it 49, 51, 59, 60 sagu 49, 58, 59 duga, -ir 64, 65, 101 sceggia (OSw.) 58 hebita, gihebit 49 salbi5e, -es 51 3

dugh~r (OSw.) 60 23 scegh~r (OSw.) 58 hlinen 73 salban, -at 52 dvína, dvena 73 sagor 49 hocta 60, 77 segis(t), -it 49, 51, 58, 59, 60 fahide (R~ln.) 77 72 seg (impv.) 49, 59 hogela 77 segita, gisegit 49 fer(r) 51 seg(i) (1 sg.) 49, 58 ktmnen 67, 73 sm, stant 97 flóa 77 71 seg(i)r 49,51,58 leben 34, 49, 72 s(w)orga 74 fl'!Íg(1') 51 segja (inf.) 49, 51, 53 libit 49, 51, 60 s(w)orgen 74 folge1' (OSw.) 66 segja (3 pl.) 49, 51 libita, gilibit 49 tagen 50 ftdger 66 segjt¿m 49 liggu, ligit 9 toug, tugem 64, 101 fylgja 66 sighia (OSw.) 58 limen, lemen 73 tugund 75 graf 59 skjót 59 magen 67, 70 wahhen 50 haf 49,59 skorta 50 meran 74 wihhan 33 hafa (inf.) 49, 51 sorg 74 mihhilan 74 -wizzen, -el 50,67,69,101 hafa (3 pl.) 7, 55, 59 talgidai (Rtm.) 7876 jir-monen 50, 63, 67 wonen 50, 72 hafa, -i(r) (opt.) 75 tawide (Run.) 77 72

hafe (ONorw.) 77 72 tel59 Old Saxon hafio 59 pmja 70 47

hafoa 77 pegi 49 folgoiad 51 ladoian 51 hafor, hafat 49, 77 pegja 49

hanga 72 pjá 74 folgon 66 libbian 34, 49, 51 hav~r (OSw.) 59 pola, -ir 50, 72 giht¿gd 56 libbiod 49 hef(i) 7, 49, 51, 53, 59 pol~r (OSw.) 60 23 ginon 73 libbit¡ 60

hef(i)1' 49, 53, 59 tma 50,72 habad, -ed 51, 84 libda 60,77 hqft¿m 59 vaka 49, 50, 51, 53 habas, -es 51 libod, lebot 49, 57 hugat 77 vaki (impv.) 59, 75 habbiad 49, 59 mornon 73

ht¿gpi (OG1ttn.) 77 vaki (1 sg.) 51 habbian 49, 59 sagda, gisagd 56, 60, 77 líoa 32 vaki1' 51 habbit¿ 49, 59 sagi 49 lifa 34, 72 vakio 51 habda, gihabd 56, 60, 77 sagis, -id 49, 58, 59 lifir 60 vakna 73 habi 49 seggiad 49, 59 liv~r (OSw.) 60 vakOa, vakat 51 habis, -id 49, 59 seggian 49, 59 livir (OSw.) 60 valda 77 72 hebbiad 7, 11, 52, 55, 84 seggit¿ 59

moma 73 vqkt¿m, vaka (3 pl.) 61 hebbian 51 segis 49 óask 67, 77 71 witadahalaiban (Run.) 76 hebbit¿ 7,11,12,51,52 tharbon 67

hogda 60,77

Old High German Old English

beben, biben 50, 62 erien 55 bi1·tl 4444 fasten 50, 74 bijian 50, 62 dw'ínan 73 blinden 50 folgen 66 blawan 37 fol3ian 66 chinan 73 ginen 73 blindian 50 fol3iaP 51 dagen 50, 63, 74 habe (impv.) 75 ceowan 20 ftdl eode 66 39 daMa 52 7 habe, -es (opt.) 75 cinan 73 fyl3(e)an 66 da'l'ben 67 habee, -es 51 3

dOO3, dt¿3an 64 3inian 73 dolen 50,72 haben, -el 7,49,52 dttgup 75 habban 51, 59

138 139

Page 70: Jasanoff.1978.Stative and Middle in Indo-European

habbap 7,55,59,84 heebbe 7, 11, 51, 59 heefde, (3e)heefd 77 heefes(t), -ep 51 heefst, -p 49, 51, 84 hafa 49 hafas(t), -ap 49 hafo 49,59 hebbe 59 22

hlaford 76 hlinian 73 h03de 77 hyc3an 49 lalJian 51 leofas(t),-ap 49 leornian 73 libban 34, 49, 51 libbap 60 Ubbe 60 lif3an 49

akfjs 86 akytas 86 16

aria 55,84 av¡s 83 be'gti 90 blikseli, -i 81 blizgeli 37 30

bl'ólis (brólys) 83 budeJo 102 budeli, -i 9, 81, 95 b1isiant- 88 busti, buñda 73 cepseli, -i 81 degü, 30 déveli, -i 90 dévime 90 dtlOS 87, 88 dúosi (2 sg,) 88 18, 89 dúosi (3 p,) 88 19

dúosiq,s, dúosiant- 87, 88 dúosime, -ite 88, 89 dúosint- 88 17

dúosiu 87, 88, 89 dúosme 89 dúoti 87 e'sti 90

140

lif3ap 49, 51 lif3e(n) 51 lif3ende 51 lifo 49 liofas, -ap 49, 57 ma3ian 67, 70 rarian 62 see,30 49, 58 sa3as(t), -ap 49 scortian 50 sealfian 51 sec3an 49,59 sec3ap59 sec3e 59 sor3ian 74 peowian 74 polian 72 w'ican 33 be-witian 67 wunian 72

Lithuanian

gaidfjs 83 gailelis, -is 74 gareli, -i 10, 85, 90 girdeli, -i 81 guleJo 102 guljJi, -i 19, 33, 83, 85, 95 gulti 33 guzeli, -a, -i, -eJa 103, 104 guzti 103 gvallCias 33 iria 84 judeli 32 juodeli, -eJa 6 júosti, -ia 83 kepu 30 líekmi, -ti 2, 91 lieziate, -ta 42 lietiü, 9 liezti, -ia 9, 42, 44 limpu 34 lipa 100 4

lüzeli, -i 83 mérdeti, -i 81, 90 27

mérdmi 90 28

myleli, -i 81 mineJo 7

e 2iIfI1

I mineU 7, 50,53, 94 mini 7, 83, 85 te-minfj 98 te-minie 98 minime, -te 83, 85 minis, minint- 82, 87, 98 miniü 9, 53, 82, 84 níedéti 90 pabégeli, -i 81 palükeli, -i 81 pavydeli, -vydi 50, 90 sakyti 58, 59 sédeli, -i 97 se'dint- 97 41 '

se'dmi 97 39

seneU, -eJa 6, 50, 81 sérgéti, sérgti 90 se'sti 97 syjéli, -i 97 syksteli, -i 74 stoveli, -i 2, 3, 85, 90, 91 stóvime 90

avs 83 bralis 83 gul', guUm 10, 85

bude 81 kil'dil, k'irdimai 81 mile 81 postasei 88 tebbei 85

stóvmi 90 28

strakseli 81 s1/,pa 1004

svérdéti, -i 81 tvitéli, svitu 95 tvyteli, -i 95 tyleli, -i 50, 83, 95 titpti, telp1i, 38 hu'eli, -i 1, 81, 83 ve"sim, -it 88 ve"sme, -te 88 veda 41, 42, 44 vedate, -ta vedt/' 41, 4444

vedúos 4444

veizdeli, -i 90 veizdmi 90 vefkti, -ia 83 -vydmi 90 28

zydeli, zydi 90 zydmi 90 28

Latvian

n'ídet, n'ist 90 raudim, -it 90 27

ziedet 90

Old Prussian

tUl''[t, t1l1'-r¡: 81, 85, 93 turrimai 10, 85 waidimai, -ti 89, 90 waidmai, waisei 89 waist 89,90

Slavic (Old Church Slavonic unmarked)

bézati 90 bogatéti, -éjet1> 6 bojati Sl!, -it1> 2, 90, 91 boléti, -it1> 90 b1>dé 102 b1>déti, -it1> 9, 95

v1>z-b1>nqti, -netb 73 bysl!st-, -qst- 88, 89 dadite 98 daZdb 98 de7aax1> 104 10

délati 104 10

de7ax1> 65 37

141

Page 71: Jasanoff.1978.Stative and Middle in Indo-European

~-~---------------------------------...,.-.,~----------------~-~~----~-~-~-~-~-~~--

r drr,zati, -it'b 83, 95 goreti, -it'b 10, 85, 90 gorQst-, -~st- 90 goSéQ, gostisi 86 gostiti, -U'b 86 gmzdane 86 jasti, jast'b 90 kosti 86 kypeti, -it'b 97 leía 102 lez'ati, -it 80, 95 pri-lbpe 34 pri-lbpeti, -it'b 19, 34, 95 pri-lb(p)nQti 34 lizati, lizet'b 9, 83 lizem'b, -ve 42 lizo 9 meljQ 38 mbne7,102 mbneaX'b, -ease 102, 104 10

mbn~st- 82, 98 mbneti 7,50,53,86,94,104 10

mbnfit'b 82 mbni, -ite (impv.) 82, 98 mbnisi 93 mbnit'b 7,83 mbnJQ 9, 53, 82, 84 u-mbrg, -et'b 82 neseax'b 104 10

nosi (aor.) 86 16

nos~st- 98 nos~t'b 94, 98 nMi (S.-O.) 86 nosim'b, -te 93 nositi, it'b 10, 83, 86, 93 nosQ, nosisi 83, 94 or"jet'b 55, 84 padom'b, -ve 42 pekQ 30 pisatí, piset'b 83 poleti, -it'b 90

142

pQtbje, pQti 86 pl'obysucny (GOz.) 88 prositi, -it'b 83, 86 Pl'OsQ, pl'Osisi 83 pytajQ 32 rr,det'¡ s~, -ejet'b (R.Oh.Sl.) 6, 50 sede 97 sed~st- 97 41

sede ti (GOz.) 97 sedeti, -it'b 97 sbcati, sbcit'b 96 slysati, -it'b 80 soCití 58 18

sociti (S.Oh.Sl.) 58, 59 stareax'b, -ease 102 stareti, -ejet'b 50, 81 stojati, -it'b 97, 99 s'bpati, -it'b 96 8Vbteti s~, -it'b 95 synove 86 S'b-tbl'et'b 84 tri 86 15

trbje 86 trr,peti, -it'b 95 t'b 93 31

umeti, -ejet'b 6 vede 2, 85, 89 vedeax'b 102 vedem'b, -ve 42 vedq, -qt'b 82, 102 vedQst- 82 vem'b 90 veste, ved~t'b 90 videti, -it'b 50, 90 vidom'b 90 28

vise ti , -it'b 83 vizdb 90, 98 42

xosti, xostesi 9842

zegq 30 zl'bdeti, -it'b (S.Oh.Sl.) 9 znal, -em, -esz (Pol.) 104

I

INNSBRUCKER BEITRAGE ZUR SPRACHWISSENSCHAFT

Publications in General Linguistics

Proceedings 01 the XIItl~ International Congress 01 Lingllists, Vienna, August 28-September 2, 1977. Edited by WOLFGANG U. DRESSLER and WOLFGANG MEID. 1979. ca. 850pp.; ca. AS 1.600.-

ELISE RICHTER: Kleinere Schriften zllr Allgemeinen Ilnd Romanischen Sprachwissen­schaft. Ausgewahlt, eingeleitet und kommentiert von YAKOV MALKIEL. Mit einer Bibliographie von B. M. WOODBRIDGE, JR. Gesamtredaktion: W. MEID. 1977. 599pp..; AS 960.-

Phonologica 1976. Akten del' 3.Internationalen Phonologie-Tagwlg, Wien, 1.-4. September 1976. Herausgegeben von WOLFGANG U. DRESSLER und OSKAR E. PFEIFFER. 1977. 31Opp.; AS 520.-

Textlingllistik und Semantik. Akten del' 4. Arbeitstagung Osterreichischer Linguisten, Innsbruck, 6.-8. Dezember 1975. Herausgegeben von WOLFGANG MEID und KARIN HELLER. 1976. 283pp.; AS 240.-

Wortbildung diachron - synchron. Akten des Kolloquiums del' Sektion für Diachrone Sprachwissenschaft im Osterreichischen Linguistischen Programm, Innsbruck, 5.-6. Dezember 1975. Herausgegeben von OSWALD PANAGL. 1976. 157pp.; AS 240.-

ANGEL O DIETMAR PEER: Studien zur Wortbildung in einer "klassischen" Translor· mationsgrammatik. Die Relativsatznominalisierung im Deutschen. 1978. XII, 483pp.; AS 480.-

ALWI.N FILL: Wortdurchsichtigkeit im Englischen. Mit einer kontl'astiven Untel'­suchung del' Rolle durchsichtiger W6rter im Englischen und Deutschen del' Gegenwal't. 1979. ca. 180pp. ; ca. AS 320.-

ILSE SCHON: Neutrum und Kollektivum. Das ~Iorphem ·a im Lateinischen und Romanisehen. 1971. VI, 140 + 35pp.; AS 240.-

To be obtained from:

Institut liir Spraehwissenschalt del' Universit1it Innsbruck A·6020 INNSBRUCK, Innrain 30