Remarks on the History of the Indo-European Infinitive

University of South CarolinaScholar Commons

Faculty Publications Linguistics, Program of

1981

Remarks on the History of the Indo-EuropeanInfinitiveDorothy DisterheftUniversity of South Carolina - Columbia, [email protected]

Follow this and additional works at: https://scholarcommons.sc.edu/ling_facpub

Part of the Linguistics Commons

This Article is brought to you by the Linguistics, Program of at Scholar Commons. It has been accepted for inclusion in Faculty Publications by anauthorized administrator of Scholar Commons. For more information, please contact [email protected].

Publication InfoPublished in Folia Linguistica Historica, Volume 2, Issue 1, 1981, pages 3-34.Disterheft, D. (1981). Remarks on the History of the Indo-European Infinitive. Folia Linguistica Historica, 2(1), 3-34. DOI: 10.1515/flih.1981.2.1.3© 1981 Societas Linguistica Europaea.

https://scholarcommons.sc.edu?utm_source=scholarcommons.sc.edu%2Fling_facpub%2F24&utm_medium=PDF&utm_campaign=PDFCoverPages

https://scholarcommons.sc.edu/ling_facpub?utm_source=scholarcommons.sc.edu%2Fling_facpub%2F24&utm_medium=PDF&utm_campaign=PDFCoverPages

https://scholarcommons.sc.edu/ling?utm_source=scholarcommons.sc.edu%2Fling_facpub%2F24&utm_medium=PDF&utm_campaign=PDFCoverPages

https://scholarcommons.sc.edu/ling_facpub?utm_source=scholarcommons.sc.edu%2Fling_facpub%2F24&utm_medium=PDF&utm_campaign=PDFCoverPages

http://network.bepress.com/hgg/discipline/371?utm_source=scholarcommons.sc.edu%2Fling_facpub%2F24&utm_medium=PDF&utm_campaign=PDFCoverPages

mailto:[email protected]

Folia Linguistica Historica II/l pp. 3—Z4 © Societae Linguistica Evropaea, lt)8l

BEMARKS ON THE HISTOKYOF THE INDO-EUROPEAN INFINITIVE

DOROTHY DISTERHEFT

1. INTRODUCTION

With the exception of Indo-Iranian (Hr) and Celtic all historicalIndo-European (IE) subgroups have a morphologically distinctInfinitive. However, no single proto-form can be reconstructed forthem — only a wide array of action nouns. A good deal has beenwritten about the development of the infinitive, almost all froma morphological perspective. But the evolution of a single infinitivefrom a large number of Proto-Indo-European ( ) action nounsis primarily a syntactic development involving not only an expan-sion in the number of clause types that use infinitive äs embeddedpredicate but also change in the structure of the infinitive clauseitself (i.e. expansion of types of subject/object occurrence), changein object case, and word Order shifts. My aim here is to identifyarchaic features, both morphological and syntactic, of attested IElanguages which will give a clearer picture of what nonfinite sub-ordinating strategies were used in certain areas of the late PDElanguage Community. This discussion differs from past syntacticstudies (e.g. those of Hahn 1950, Miller 1974, Ard 1975) which focuson embedded structures of a further-developed type (e.g. thosewhere subject-raising is involved) to reconstruct earlier stagesof infinitival Subordination. I will argue here that under certainconditions the action noun was indeterminately a noun or the pre-dicate of a subordinate clause. Beanalysis äs an infinitive resultedin a complete shift in most languages to part of the verbal System.

* A preliminary sketch of this papor was presented at the Winter 1978LSA meeting. I wish to thank Carol F. Justus and Theodore M. Lightnerfor their comrnents on i t and also Sandra A. Thompson and Patrick K. Fordfor helj) duiing previous stages of this researeh. Of course, I take füll re-sponsibility for presentation and analysis of the data.

Brought to you by | University of South Carolina Libraries (University of South Carolina School of L)Authenticated | 172.16.1.226

Download Date | 5/31/12 9:18 PM

i

I I I

Foli.lAfl//ui.tfco HfltorlcG II/I 'P'P. 8-U © 800£etru lAngu,.tica Etlropaea. 1981

REMARKS ON THE HISTORY OF THE INDO-EUROPEAN INFINITIVE

DOROTHY DISTERHEFT

1. INTRODUCTION

With the exceptfon of Indo-Iranian (Ur) and Celtic all historical Indo-European (IE) subgroups have a morphologicalIy distinct infinitive. However, no single proto-form can be reconstructed for them - only a wide array of action nouns. A good deal has been written about the development of the infinitive, almost all from a morphological perspective. But the evolution of a single infinitive from a large number of Proto-Indo-European (PIE) action nouns is primarily a syntactio development involving not only an expansion in the number of clause types that use infinitive as embedded predicate but also change in the struoture of the infinitive clause itself (i.e. expansion of types of subject/object occurrence), change in object case, and word order shifts. My aim here is to identify archaic features, both morphological and syntactio, of attested IE languages whioh will give a clearer pioture of what nonfinite subordinating strategies were used in oertain areas of the late PIE language, community. This'discussion differs from paat syntaotio studies (e.g. those of Hahn 1950, MiUer 1974, Ard 1975) whioh focus on embedded structures of a further-developed type (e.g. those where subject-raising is involved) to reconstruot earlier stages of infinitival subordination. I will argue here that under oertain conditions the action noun was indeterminately a noun or the predicate of a subordinate olause. Reanalysis as an infinitive resulted in a complete shift in most ]an~ages to part of the verbal system .

... A pre~ sketch of this paper was preSented at the Winter 1978 LSA meeting. I Wish to thank Carol F. Justus and Theodore M. Lightner for their comments on it and also Sandra A. Thompson and Patrick K. Ford for help during previous stages of this research. Of course, I take fuJI responsibility for presentation and analysis of the data.

The data presented here are from the two language groups, Ilrand Celtic,1 that preserve several IE action nouns äs infinitives.I conclude that Old Irish (Olr) has subordinate clauses very closeto what had, while the Rigvedic (RV) period of Sanskrit hasinnovated from purely nominal Status; Avestan (Av), representingOld Iranian, also shows traces of earlier stages even though it hasstandardized object case selection which was variant in Ilr. Othertopics like the extent to which morphology and syntax are con-nected and what conditions were necessary to allow incipientinfinitive analyses will be explored. I will also sketch the morpho-logical and syntactic consequences of this reanalysis and comparethem with other IE languages which have more developed infinitiveclauses.

2. INDO-IRANIAN AND' CELTIC

The infinitives of RV and Av are oblique cases of productiveaction nouns with the only exceptions being RV -dhyai and Av-dyäi «Ilr *-dhyäi),RV -sani and-tavai (u). *-dhyäiis morpholog-ically obscure and not to be derived from any IE nominal para-digm,2 while RV -sani is the locative singular of an extendedheteroclite stem (-sar-l-san-; for morphological details see Ben-veniste 1935b: Chapter 1) which, though attested in other

1 My description of Celtic is based upon Old Irish which is the best attestedof the older languages in that group. Traditional descriptions of these lan-guages tend to take the position that the verbal noun is not an infinitive(e.g. Thurneysen 1946: 455, Dillon 1955: 112 f.); in this discussion I showthat in certain clause types infinitival Status should be assigned to someOld Irish structures.

2 But cf. arguments like those by Bartholomae 1890, Schwyzer 1953/1:809 which try to match *-dhyäi with the Greek middle infinitive -sthai.This correspondence is untenable on phonological grounds: Hr *-dhy-corresponds to Gk -ss- (Benveniste 1935b: 208). I have discussed elsewhere(Disterheft 1980: 34—40) earlier elaims, notably by Benveniste 1935a:Chapter 2, that -*dhyäi is a medio-passive infinitive. His two criteria are:1. if the infinitive has the same meaning äs the finite medio-passive, it hasthe same voice äs well; 2. if the finite paradigms are predominantly middleor passive the infinitive is likewise middle. My examination of the data uponwhich* Benveniste's claim is based do not support his argument: while outof a total of 33 verbal Sterns, 8 are exclusively and 4 are predominantlymiddle, 7 appear equally often in both active and middle, 11 are infrequentlymiddle, and 3 are exclusively active. Since middle voice does not changethe marking of agent's and patient's relationship to the predicate, it is oftendifficult to infer that middle is the voice to be assigned to the predicateunless it is obvious from the meaning of the verbal stem or from the restof the sentence.



4

The data presented here are from the two language groups, IIr and Celtic,l that preserve several IE action nouns as infinitives. I conclude that Old Irish (OIr) has subordinate clauses very close to what PIE had, while the Rigvedic (RV) period of Sanskrit has innovated from purely nominal status; Avestan (Av), representing Old Iranian, also shows traces of earlier stages even though it has standardized object case selection which was variant in IIr. Other topics like the extent to which morphology and syntax are connected and what conditions were necessary to allow incipient infinitive analyses will be explored. I will also sketch the morphological and syntactic consequences of this reanalysis and compare them with other IE languages which have more developed infinitive clauses.

2. INDO-IRANIAN AND' CELTIC

The infinitives of RV and A v are oblique cases of productive action nouns with the only exceptions being RV -dhyai and Av -dyai ( <IIr *-dhyai) , RV -sani and -tavai (u). *-dhyai is morphologically obscure and not to be derived from any IE nominal paradigm,2 while RV -sani is the locative singular of an extended heteroclite stem (-sar-/-san-; for morphological details see Benveniste 1935b: Chapter 1) which, though attested in other IE

1 My description of Celtic is based upon Old Irish which is the best attested of the older languages in that group. Traditional descriptions of these languages tend to take the position that the verbal noun is not an infinitive (e.g. Thurneysen 1946: 455, Dillon 1955: 112 f.); in this discussion I show that in certain clause types infinitival status should be assigned to some Old Irish structures.

2 But cf. arguments like those by Bartholomae 1890, Schwyzer 1953/1: 809 which try to match *-dhyai with the Greek middle infinitive -8thai. This correspondence is untenable on phonological grounds: IIr *-dhycorresponds to Gk -88- (Benveniste 1935b: 208). I have discussed elsewhere (Disterheft 1980: 34-40) earlier claims, notably by Benveniste 1935a: Chapter 2, that -*dhyai is a medio-passive infinitive. His two criteria are: 1. if the infinitive has the same meaning as the finite medio-passive, it has the same voice as well; 2. if the finite paradigms are predominantly middle or :p~s~ive the ~i~tiv~ is ~ikewise middle. My examir~ation of the data upon whlCn BenvenISte s claIm IS based do not support hIS argument: while out of a total of 33 verbal s.tems, 8 are exclusively and 4 are predominantly middle, 7 appear equally often in both active and middle, 11 are infrequently middle, and 3 are exclusively active. Since middle voice does not change t~e .marking. of agent's a~d pat~ent's rel~tionship to t~e predicate, it is often dlfflCul~ t.o Infe~ that mIddle IS th~ vOIce to be assigned to the predicate unless It IS obVIOUS from the mearung of the verbal stem or from the rest of the sentence.

languages, is no longer used nominally in Indic. -tavai (u) is, accord-ing to Thurneysen 1908, derived from -tave (dative -tu- stem).3RV and Av most commonly use dative, accusative, and locativecases of *-ta-, *-&*-, *-os-, root, and various extended heteroclitestems. Olr verbal nouns (the term for actioii nouns in traditionalCeltic grammar) reconstruct to basically the same inventory,though some restructuring has taken place. Dative and accusativeOlr verbal nouns appear in clauses similar to those of Ilr eventhough they are without exception members of nominal paradigms.

Since most Ilr infinitives and all Olr verbal nouns are morpho-logically identical to nouns, in certain subordinate clauses somedative and accusative action nouns may be interpreted äs eithersubordinate predicates or simple NPs. This problem of analysishas certainly beeil" recognized; differentiation of substantives frominfinitives has been an important theme in previous research withemphasis placed on morphological factors or object case selection(for literature see Disterheft 1980r28-—34). Here I shall focuson the syntactic structure of clauses in which these action nounsappear in order to determine if they are simple nouns or predicatesof subordinate clauses. Fast studies have depended upon objectcase selection äs the criterion for infinitivehood: if object is thesame case äs that of finite verb (usually accusative) the analysisof the passage in question is infinitival. Conversely, if the object(or subject) is genitive, the form is nominal. It is indeed the casethat genitive agent/patient marking is Standard with action nounsäs is accusative with "real" infinitives; but when we are dealingwith grammars that have transitional forms — äs the KV mostdefinitely is to judge by its morphology — the Situation is not sowell-defined. It can be argued that agent/patient marking is notthe only criterion. For instance, the infinitive is predicate in sen-tences that are mildly imperatival and almost always have accu-sative object. But in (1) (where the infinitive is the locative ofan extended heteroclite stem, -tar-), the object of the infinitive isgenitive even though finite forms of dty- never select that case.

3 Thurneysen 1908 suggested that -tavai with its anomalous second wprdaccent is derived from -tave (pre-Vedic *-tavai) followed by the particlevai: *-tavai vai then haplologized to its historical form. Klein 1978: 164-—7agreed on most points with Thurneysen, adding that u was added to themetrically incorrect line that resulted from the haplology. He demonstratednot only that -tavai is used in the same syntactio and metrical collocationsäs •tave» but also that it appears in slightly younger texte.



5

languages, is no longer used nominally in Indic. -tavai (u) is, according to Thurneysen 1908, derived from -tave (dative -tu- stem).3 RV and Av most commonly use dative, accusative, and locative cases of *-tu-, *-ti-, *-as-, root, and various extended heteroclite stems. OIr verbal nouns (the term for action nouns in traditional Celtic grammar) reconstruct to basically the same inventory, though some restructuring has taken place. Dative and accusative OIr verbal nouns appear in clauses similar to those of IIr even though they are without exception members of nominal paradigms.

Since most IIr infinitives and all OIr verbal nouns are morphologically identical to nouns, in certain subordinate clauses some dative and accusative action nouns may be interpreted as either subordinate pre~~ates or simple NPs. This problem of analysis has certainly beeh'recognized ; differentiation of substantives from infinitives has been an important theme in previous research with emphasis placed on morphological factors or object case selection (for literature see Disterheft 1980: '28-34). Here I shall focus on the syntactic structure of clauses in which these action nouns appear in order to determine if they are simple nouns or predicates of subordinate clauses. Past studies have depended upon object case selection as the criterion for infinitivehood: if object is the same case as that of finite verb (usually accusative) the analysis of the passage in question is infinitival. Conversely, if the object (or subject) is genitive, the form is nominal. It is indeed the case that genitive agent/patient marking is standard with action nouns as is accusative with "real" infinitives; but when we are dealing with grammars that have transitional forms - as the RV most definitely is to judge by· its morphology - the situation is not so well-defined. It can be argued that agent/patient marking is not the only criterion. For instance, the infinitive is predicate in sentences that are mildly imperatival and almost always have accusative object. But in (I) (where the infinitive is the locative of an extended heteroclite stem, -taro), the object of the infinitive is genitive even though finite forms of dl~r- never seJect that case.

3Thurneysen 1908 suggested that .tavai with its anomalous second word accent is derived from -tave (pre-Vedic *-tavai) followed by the particle vdi: *-tavai vai then haplologized to its historical form. Klein 1978: 164-7 agreed on most points with Thurneysen, adding that 'It was added to the metrically incorrect line that resulted from the haplology. He demonstl'l!'ted not only that -tavai is used in the same syntactic and metrical collocatIOns as -tave. but also that it appears in slightly younger texts,

6

(1) aa . . . brahmanaspatir druho hantä maha rtasya dhartarithis Brahmanaspati Lie slayer great Truth upholdnom nom gen nom gen gen inf

(II. 23.17)'May this Brahmanaspati, the slayer of the Lie. uphold thegreat Truth/

Thus we see that the single feature of ob ject case selection is inade-quate for purposes of defining infinitives, äs Renou 1954 has alsorecognized. I propose here that more properties of the clause mustbe used to define Infinitive Status, at least in the transitionalInfinitive grammars of Ilr and Celtic. A more reliable approachis to include the treatment of subordinate subject :4 if it has thesame properties that subordinate infinitive subject does in lan-guages with morphologically separate infinitives, the action nounin question should be considered infinitival. Major processes whichsubjects of infinitives undergo are:

a. equivalent noun phrase deletion ;b. raising to object;c. raising to subject.

When subordinate subject is not coreferent with a main clause NPwe also find:

d. overt subject in dative or accusative case.\

In 2.1—2.3 the features of purpose clauses and verb complementsin RV, Av, and Olr will be described and then compared to simpleobject/oblique case nouns. Action nouns which are indeterminatelyinfinitives/substantives will also be introduced and contrasted withforms known to be infinitives.

2.1 INDIO. The nonfinite purpose clause in RV uses an infinitiveäs predicate with no conjunction introducing it. Subject is eitherdeleted by coreference with an NP in the main clause (subject,direct object, indirect object, possessive, prepositional phrase) oris an overt dative. In (2), the -dhyai infinitive has accusative

4 Sentences with an infinitive subject like Lat errare humanum est donot occur in languages with nominally associated forms. Sentential subjectswith an infinitive typically appear in grammars where the infinitive hasbeen separated from the nominal paradigm for so long that it retains noparadigmatic association, such äs case marking, which would keep obliquecases from appearing in subject slots.



6

(1) sa ... brahma'J')a8patir druM hanta maM '{tMya dhartari this Brahma~aspati Lie slayer great Truth uphold nom nom gen nom gen gen inf

(II. 23.17)

'May this Brahma~aspati, the slayer of the Lie. uphold the great Truth:

Thus we see that the single feature of object case selection is inadequate for purposes of defining infinitives, as Renou 1954 has also recognized. J propose here that more properties of the clause must be used to define infinitive status, at least in the transitional infinitive grammars of IIr and Celtic. A more reliable approach is to include the treatment of subordinate subject:4 if it has the same properties that subordinate infinitive subject does in languages with morphologically separate ~initives, the action noun in question should be considered infinitival. Major processes which subjects of infinitives undergo are:

a. equivalent noun phrase deletion; b. raising to object; c. raising to subject.

When subordinate subject is not coreferent with a main clause NP we also find:

d. overt subject in dative or accusative case.

In 2.1-2.3 the features of purpose clauses and verb complements in RV, Av, and OJr will be described and ·then compared to simple object/oblique case nouns. Action nouns which are indeterminately infinitives/substantives will also be introduced and contrasted with forms known to be infinitives.

'-2.1 INDIC. The nonfinite purpose clause in RV uses an infinitive

as predicate with no conjunction introducing it. Subject is either deleted by coreference with an NP in the main clause (subject, direct object,'. indirect object, possessive, prepositional phrase) or is an overt dative. In (2), the -dhyai infinitive has accusative

4 Sentences with an infinitive subject like Lat errare humanum est do not occur in langua.ges with nominally associa.ted forms. Sentential subjects with an infinitive typically appear in grammars where the infinitive has been ~eparB:ted fro~ ~he nominal paradigm. for so ~ong that it retains no paradIgmatIC aSSOCIatIOn, such as case marking, which would keep oblique cases from appearing in subject slots. . .

object with subject deleted by coreference with main clause sub-ject.

(2) td vigram dJiaithe jatharam pfnadhyai (VI. 67.7)DEM strong take stomach fillnom du acc 2 du imv acc inf'You both should take the strong one in order to fill yourstomach/

In (3) the -tavai (u) infinitive has subject deleted by coreferencewith the object of the main clause verb. Its dative object is commonin this clause type.

<' cx

(3) brahmdna Indram mdhayanto arkair avardhayannBrahmans Indra magnifying songs they-have-strengthenednom acc nom pl insträhaye MAtava u (V. 31.4)snake killdat inf'The magnifying Brahmans have strengthened Indra withsongs so that (he) would slay the snake/

Even though KV has no morphological voice distinction, passiviza-tion is indicated in all clause types by marking patient äs subject.(4) has an infinitive whose subject is deleted by coreference withmatrix indirect object.

(4) sa na a vdha ... rayim ..„ isayadhyai (VI. 64.4)DEM us P bring wealth strengthennom sg encl 2 sg imv acc inf'Bring wealth to us ... that (we) may be strengthened/

Nominal infinitives are predicates in clauses identical to the non-nominal ones of (2)—(4). (5) ütaye is a member of a paradigm.(dative -tistem) attested in most cases of both Singular and pluralIt is listed in Grassmann's Wörterbuch zum Rigveda (1873) onlyunder the noun üti- 'help*. However, this dative action noun isused äs an infinitival predicate in an embedded clause (subjectdeleted by coreference with matrix object, accusative object):



7

object with subject deleted by coreference with main clause subject.

(2) ta wgram dkaitke ialMram 'Pr'(Uidkyai (VI. 67.7) DEM strong take stomach fill nom du acc 2 du imv acc inf 'You both should take the strong one in order to fill your stomach.'

In (3) the -tavai ('II.) infinitive has subject deleted by coreference with the object of the main clause verb. Its dative object is common in this clause type.

(( (3) brrikmii'(l.a iMram maMyanto arlctiir o,vardkayann

Brahmans Indra magnifying songs they-have-strengthened nom acc nom pI instr anaye Mnta'Vit 'II. (V. 31.4) snake kill dat inf 'The magnifying Brahmans have strengthened Indra with songs so that (he) would slay the snake:

Even though RV has no morphological voice distinction, passivization is indicated in all clause types by marking patient as subject. (4) has an infinitive whose subject is deleted by coreference with matrix indirect object.

(4) 8it na it VaM . • 0. rayim ..• i,ayadkyai (VI. 64.4) DEM us P bring wealth strengthen nom sg encl 2 sg imvacc inf 'Bring wealth to us ... that .(we) may be strengthened:

Nominal infinitives are predicates in clauses identical to the nonnominal ones of (2}-(4). (5) iU6.ye is a member of a paradigm. (dative -tistem) attested in most cases of both singular and plural It is listed in Grassmann's Worlerbuch zum Rigveda (1873) only .under the noun uti- 'help'. However, this dative action noun is used as an infinitival predicate in an embedded clause (subject deleted by coreference with matrix object, accusative object):

8

(5) eva vasiftha indram ütaye nfn ... gynäti (VII. 26.5)therefore Vasishtha Indra help men he-praises

nom acc inf acc'Therefore, Vasishtha praises Indra in order that (he) helpthe men/

vipfccharh 'inquire' (6) is the accusative of a root noun in a purposeclause and, like isayadhyai in (4), is the only constituent in itsclause to appear in the sentence. The root pfch- is rarely nominal-ized and is found only in the dative and accusative cases ; in thispassage preverb vi + accusative action noun can only be inter-preted infinitivally.

(6) upo emi cikituso vipfccham (VII. 86.3)towards I-go wise-ones inquire

acc infapproach the wise ones in order to inquire/

Likewise the dative of a productive stem may also be predicatein a clause where it is the only constituent. oubhe 'shine' (7) is thedative of a root noun which is attested in a füll paradigm in thesingular.

(7) samänam angy änjate subhe kamsame ointment theyânoint-themselves shine PTacc acc inf (VII. 57.3)'They anoint themselves in the same ointment in order to shine.'

(8) and (9) illustrate dative subject with dative infirdtives formedon root stems: (8) is intransitive; dfoe (9) has accusative object.Like ütaye in (5), vfdM is listed in"the dictionary under the nominalparadigm. d?se, however, is found äs an Infinitive under the verbalroot - and has no extant nominal paradigm.

(8) dbhi gandharvam atynad ahudhnösu rajdhsvto Gandharva he-crossed bottomless skies

acc - loc locä indro brahmtähya id vfdha (VIII. 77.5)P Indra Brahmans PT prosper

nom dat inf



8

(5) eva vll8i~/ha Uulram fUage nfn . .. gr1]iiti (VII. 26.5) therefore Vasishtha Indra help men he-praises

nom acc inf acc 'Therefore, Vasishtha praises Indra in order that (he) help the men:

vipfccharh 'inquire' (6) is the accusative of a root noun in a purpose clause and, like i~agQ,dhgai in (4), is the only constituent in its clause to appear in the sentence. The root prch- is rarely nominalized and is found only in the dative and accusative cases; in this passage preverb vi + accusative action noun can only be interpreted infinitivally.

(6) upo emi cikit~o vipfccham (VII. 86.3) towards I -go wise-ones inquire

acc inf 'I approach the wise ones in order to inquire:

Likewise the dative of a productive stem may also be predicate in a clause where it is the only constituent. subM'shine' (7) is the dative of a root noun which is attested in a full paradigm in the singular.

(7) samiinam angy anjate subM him same ointment they-anoint-themselves'shine PT acc acc inf (VII. 57.3) 'They anoint themselves in the same ointment in order to shine. '

(8) and (9) illustrate dative subject with dative infinitives formed on root stems: vrdM (8) is intransitive; drse (9) has accusative object. Like utaye in (5), vrdM is listed in-the dictionary under the nominal paradigm. drse, however, is found as an infinitive under the verbal rootdts- and has no extant nominal paradigm.

(8) abhi gandharvam atr'l)ad abudhne§u rajaltsv to Gandharva he-crossed bottomless skies

acc loc loc a indro brahmabhya id V'{dM (VIII. 77.5) P Indra Brahmans PT prosper . nom dat inf

9

'Indra crossed to the Gandharva in the bottomless skies inorder that the Brahmans might prosper/

(9) ud u tyam jätavedasam devam vahaniitherePTthis Jatavedasa god they-carry

acc acc accketavah vtäväya suryam (L 50.1)rays see all sunnom inf dat acc'The rays carry the god Jatavedasa there in order that allmay see the sun/

In the KV, infinitive complements are formed to verbs whosesubjects control equi deletion of infinitive subject.5 (10) has thedative samdfse 'see' with accusative object.

(10) kavtnr icchämi samdföe sumedhah (III. 38.1)poets I-wish see wise-oneacc inf nom

, the wise one, wish to see the poets/

In (11) the complement infinitive is an accusative which alsotakes accusative object:

(11) oakema väjino yamam (II. 5.1)may-we-be-able war-horses leadopt acc inf'May we be able to lead the war horses/

(12) has accusative ärdbham 'be allied with^ äs an intransitive:

(12) na pancabhir duabhir va$fy äraJbham (V. 34.5)NEG five ten he-wishes be-allied

instr instr inf'He does not wish to be allied with five (or) with ten (men)/

* KV verbal roots in this categoiy are -, i$- 'wish', jut- 'like, desire',jnä-, cü-, vid- 4know', man- 'think', ^ /c-, arh-, $- *be, able*, arthaya- 'stive',arhs- *succeed*.



9

'Indra crossed to the Gandharva in the bottomless skies in order that the Brahmans might prosper:

(9) -ad u tyarh jatavedasarh devarh vahanti therePTthis Jatavedasa god. they-carry

acc acc acc ketavalb dr8e v£8vaya suryarh (I. 50.1) rays see all sun nom inf dat acc 'The rays ·carry the god Jatavedasa there in order that all may see the sun:

In the RV, infinitive complements are formed to verbs whose ./

subjects control equi deletion of infinitive subject.5 (10) has the dative sarhdfSe 'see' with accusative object.

(10) kavtiir icchiimi sarhdfse sumedMl} (III. 38.1) poets I-wish see wise-one acc inf nom 'I, the wise one, wish to see the poets:

In (11) the complement infinitive is an accusative which also takes accusative object:

(U) sakema vaj£no yamarh (II. 5.1) may-we-be-able war-horses lead opt acc inf 'May we be able to lead the war horses.'

(12) has accusative arabham 'be allied with' as an intransitive:

(12) na paiu;ti,bhir dMhhir va§ly artibharh (V. 34.5) NEG five ten he-wishes be-allied

instr instr inf 'He does not wish to be allied with five (or) with ten (men):

II RV verbal roots in this category ara vaS-, i~- 'wish', ju~- 'like, deRire', jna-, cit-, vid- 'know', man- 'think', Aak-, arn-, i~- 'be, able', arthaya- 'st,iva', am8- ·succeed'.

10

Compleraents to mäh may also have overt, noncoreferent subjectin the dative case:

(13) yathä ta u&maei istaye (L 30.12)when you we-wish hasten

dat inf'when we wish that you hasten'

In the above examples illustrating purpose clauses and verbcomplements, a combination of features leads to the determinationthat these are Infinitive clauses. Either the overt presence of asubject (8), (9), (13) or a dative (3) or accusative (2), (5), (9), 10),(11) object gives syntactic confirmation to analysis äs infinitives.Likewise, a nonnominal infinitive such äs (4) isayadhyai guaranteesthe same Interpretation even though it is the only surface con-stituent of the clause. The accusatives vipfccham (6) and ärdbham(12) are the only members of clauses after approach and wish,respectively, and thus could conceivably be analyzed äs directobjects. However, they belong to prefixed stems which are attestedonly in infinitival cases (dative, accusative). subM (7) is the dativeof a well-established paradigm in a sentence which would not makesense if analyzed nomnfinitivally.

Parallel to the infinitival accusatives of (6), (11), (12) are formsthat can only be interpreted äs direct objects op accusatives of goalafter verbs of motion. For example, in (14) pitim ' drink' (accusative-tistem) expresses the goal of yahi 'come':

(14) ugrasv in nu. . . trikadrukesu pahi somam indra ...powerful PT now Trikadrukas drink soma Indraloc loc 2 sg imv acc vocyahi haribhyäm sutasya . pitim (II. 11.17)come golden-ones extract drink2 sg imv instr gen acc'Now, o Indra,.. . drink the soma among the powerful Tri-kadrukas . . . come with the golden ones to the drinking of theextract/

Here the accusative action noun is certainly substantival: there isno evidence that a coreferential infinitive subject has been deletedand the meaning of the entire stanza mitigates against interpreting



10

Complements to wi8h may also have overt, noncoreferent subject in the dative CMe:

(13) yatha ta uSmasi il}lUye (1. 30.12) when you we-wish hasten

dat inf 'when we wish that you hasten'

In the above examples illustrating purpose clauses and verb complements, a combination of features leads to the determination that these are infinitive clauses. Either the overt presence of a subject (8), (9), (13) or a dative (3) or accusative (2), (5), (9), 10), (11) ol)ject gives syntactic confirmation to analysis as infinitives. Likewise, a nonnominal infinitive such as (4) i~ayadhyai guarantees the same interpretation even though it ,is the only surface constituent of the clause. The accusatives viptccham (6) and arabham (12) are the only members of clauses after approach and wi8h, respectively, and thus could conceivably be analyzed as direct objects. However, they belong to prefixed stems which are attested only in infinitival cases (dative, accusative). 8ubM (7) is the dative of a well-established paradigm in a sentence which would not make sense if analyzed noninfinitivally.

Parallel to the infinitival accusatives of (6), (11), (12) are forms that can only be interpreted as direct objects of accusatives of goal after verbs of motion. For example, in (14) pitim 'drink' (accusative -tistem) expresses the goal of yahi 'come' :

(14) ugre~v in nu ... trf,kadruke~u pahi 86mam iiulra • •• powerful PT now Trikadrukas drink soma Indra loc loc.. 2 sg imv acc voc yahi haribhyarh 8utasya. piUm (II. 11.17) come golden-ones extract drink 2 sg imv instr gen acc 'Now, 0 Indra, ... drink the soma among the powerful Trikadrukas ... come with the golden ones to the drinking of the extract.'

Here the accusative action noun is certainly substantival: there is no evidence that a coreferential infinitive subject has been deleted and the meaning of the entire stanza mi~igates against interpreting

11

pttim s an accusative infinitive. y hi . « . p im is struoturally parallelto pahi ... somam. Thus, in 'drink the soina . . . coine to thodrinking* the two accusatives (the seoond derived from a verbalroot, the first one not) are parallel to eaoh other. Note that theconstituents of (14) (accusative action noun following a verb ofmotion) are very close to (6) upo &mi... vipfccJiam Ί approaohin order to inquire'. In (15) the action noun ( tim, accusative -Bi-stern) is the direct object of *ask':

(15) vfsantamasya Ti mahe tim (L 10.10)most-manly we-ask helpgen acc'We ask the help of the most manly one/

'. <

Dative substantives likewise appear in sentences siinilar to thosewhere the dative has been analyzed s a purposive infinitive (5),(7), (8), (9). For example, the dative root noun έηΜιέ (16) is identicalto that of (7).

(16) vaksahsu rukmdfi adhi yetire oubho (L 64.4)breasts gold upon they-bind adornment0loc acc pl dat'They bind the gold upon their breasts s an adornment/

In contrast to the action nouns which are infinitival (2)—(13)or nominal (14)—(16), there is a large group of datives and accu-satives whose Status is difficult to establish. In addition to aninfinitive Interpretation, the dative forms may be analyzed snouns expressing the goal of an action while the accusatives maybe either the direct object of a transitive verb or the goal of amotion verb. For example, subM (17), unlike the infinitive of (7)or the Substantive of (16), has an analysis which is indeterminate.

(17) ένύΗέ kam y nti rathatarbhir a vaih (I. 88.2)glorify/glory PT they-go wagon-pulling horsesinf dat NP instr pl instr

β The difference between tho gloas of aubho here and in (7) w thoof changes in transitivity: έυώΗ- i translated a 'ehino, glow* when intmn·itive, but *adore, glorify* when transitive.



11

pitim, as an accusative infinitive. yahi ••. piti1n is struoturally parallel to pii,hi ••• somam. Thus, in 'drink the soma ... come to the drinking' the two aoousatives (the second derived from a verbal root, the first one not) are parallel to each other. Note that the oonstituents of (14) (acousative action noun following a verb of motion) are very close to (6) upo emi . .. 'Ilipfccham 'I approaoh in order to inquire'. In (I5) the action noun (utim. accusative -tistem) is the direct object of 'ask' :

(15) vf~antamasya humahe 'l'Ui.rh (I. 10.10) most-manly we-ask help gell acc 'We ask the help of the most manly one.'

I. "

Dative substantives likewise appear in sentences similar to those where the dative has been analyzed as a purposive infinitive (5), (7), (8), (9). For example, the dative root noun 8ubM (16) is identical to that of (7).

(16) vak§altsu rulcmii,ii, adhi yetire SubM (1. 64.4) breasts gold upon they-bind adornment' 100 aco pI dat 'They bind the gold upon their breasts as an adornment.'

In contrast to the action nouns which are infinitival (2)-(13) or nominal (14)-(16), there is a large group of datives and accusatives whose status is difficult to establish. In addition to an infinitive interpretation, the dative forms may be analyzed as nouns expressing the goal of an action while the accusatives may be either the direct object of a transitive verb or the goal of a motion verb. For example, 800M (17), unlike the infinitive of (7) or the substantive of (16), has an analysis which is indeterminate.

(17) Au1JM kdrh yanti rathat1trbhir a8vai(/, (1. 88.2) glorify/glory PT they-go wagon-pulling horses inf dat NP instr pI instr

• The difference between Lhe gl08f\ of BubhG hera and in (7) iH tho, rlll'ult of changes in transitivity; Aubh· is tra.nsla.ted a.s 'shine, glow' whon IOt.rull· sitive, but 'adore, glorify' when tra.nsitivG.

I

12

dative of goal: 'They go with the wagon-pulling horses forglory/Purposive Infinitive: 'They go with the wagon-pulling horsesin order to be glorified/

ütaye (18) is the same action noun found in the purpose clauseof (5). Here it has no object and appears after invoke, äs does theSubstantive of (15).

(18) indram a Tiuva ütaye (L 111.4)Indra P I-invoke helpacc dat NP/inf

Its dative case and the meaning of the entire sentence yieldthe following interpretations:

noun: invoke Indra for help/infinitive: invoke Indra in order that (he) might help/Action nouns after wish, desire likewise may be ambiguous since

the RV allows both nominal and sentential objects with it. In (19)the meaning of the action noun gätum can be extended from 'agoing' to 'path' and cooccurs with the dative name manave:

(19) aträ däsasya namuceh oiro yad avartaypthen demon Namuci head when you-caused-to-turn

gen gen accmanave gätum icchan (V. 30.7)Manu go/path desiringdat inf acc NP nom sg ptc

Since gätum is accusative, it can be interpreted either äs the objectof desire ('at the time when you caused the head of the demonNamuci to turn, desiring a path for Manu') or its complement withdative subject ('. . .desiring that Manu go'). The similar clause in(13) differs from (19) by having a dative action noun which allowsonly a nonobjective, infinitival Interpretation.

The discussion so fär has shown that in the RV:a. Infinitives in verb complements and purpose clauses appear

with both nonnominal (-dhyai, -sani, -tavai (u)) and nominal(datives, accusatives of productive action nouns) forms;



12

dative of goal: 'They go with the wagon-pulling horses for glory.' Purposive infinitive: 'They go with the wagon-pulling horses in order to be glorified.'

Ut4ye (18) is the same action noun found in the purpose clause of (5). Here it has no object and appears after invoke, as does the substantive of (15).

(18) iiulrdm Ii huva utaye (I. 111.4) Indra P I-invoke help acc dat NP/inf

Its dative case and the meaning of the entire sentence yield the following interpretations:

noun: 'I invoke Indra for help.' infinitive: 'I invoke Indra in order that (he) might help.' Action nouns after wish, desire likewise may be ambiguous since

the RV allows both nominal and sentential objects with it. In (19) the meaning of the action nouI!. gatUm can be extended from 'a going' to 'path' and cooccurs with the dative name manave:

(19) 6;tra dasasya ndm'UC~ 8iro ydd avartayp then demon Namuci. head when you-caused-to-turn

gen gen acc manave gatum icchdn (V. 30.7) Manu go/path desiring dat . inf acc NP nom sg ptc

Since gatum is accusative, it can be interpreted either as the object of desire ('at the time when you caused the head of the demon Namuci to turn, desiring a path for Manu') or its complement with dative subject (' ... desiring that Manu go'). The similar clause in (13) differs from (19) by having a dative action noun which allows only a nonobjective, infinitival interpretation.

The discussion so far has shown that in the RV : a. Infinitives in verb complements and purpose clauses appear

with both nonnominal(-dhyai, -sani, -tavai (u)} and nominal (datives, accusatives of productive a~tion nouns) forms;

13

b. All nonnominal forms (2)-—(4) have exclusively infinitivalinterpretations;

c. Accusatives of productive paradigms are most often used äscomplements to subject-equi verbs (10)—(12) and may be analyzedindeterminately äs complements to or direct objects of a finiteverb (19);

d. Datives of productive paradigms are predominantly purposiveinfinitives (5), (7), (8), (9) but some are indeterminately predicatesin purpose clauses or dative nouns (17), (18);

e. Productive accusatives and datives which are only nominal(direct object (14), (15) or dative of goal (16), respectively) preservethe earliest stage in which these had only nominal properties ;

f. Datives whicly<are used äs complements to subject-equi verbs(13) and accusatives äs purposive infinitives (6) are never indeter-minate because they have case marking which does not allow themto be confused with nouns in their respective positions; theyshould be considered a later development than stage (e) but earlierthan that of (b).

2.2. IRANIAN. The clause types and syntactic details of theAv infinitives are almost identical to the BV. The most strikingdifference is that accusative object has been generalized by thetime that the texts were recorded (20).

(20)—(22) illustrate purpose clauses which regularly have nosubordinating marker. The subjects of nonnominal -dyäi (Ilr*-dhyäi) and the datives davoi, anäSe are deleted by coreferencewith main clause subject:

(20) kaßä aSäi druföm dyqm zastayo nl hlmhow Truth Lie I-would-place hand down it

dat acc loc accm&rqzdyäi dwahyä mqdräis s§nghahyä vmavaiilm sinqm dävoicast-down your words teaching powerful chasm giveinf instr gen acc acc infdrdgvasü ä IS dväßzng mazdä qstqs- cä(Y44.14)Liars to them suffering Wise bring hostility andloc acc acc voc inf acc'How would I place the Lie in the hand of Truth in order tocast it down by the words of your teaching, in order to put



13

b. All nonnominal forms (2)-(4) have exclusively infinit,ival interpretations ;

c. Accusatives of productive paradigms are most often used as complements to subject-equi verbs (10.)-(12) and may be analyzed indeterminately as complements to or direct objects of a finite verb (19);

d. Datives of productive paradigmS are predominantly purposive infinitives (5), (7), (8), (9) but SOme are iildeterminately predicates in purpose clauses or dative nouns (17), (18) ;

e. Productive accusatives and datives which are only nominal (direct object (14), (15) or dative of goal (16), respectively) preserve the earliest stage in which these had only nominal properties;

f. Datives whic:Q:(are used as complements to subject-equi verbs (13) and accusatives as purposive infinitives (6) are never indeterminate because they have case marking which does not allow them to be confused with nouns in the~. respective positions; they should be considered a later development than stage (e) but earlier than that of (b).

2.2. IRANIAN. The clause types and syntactic details of the Av infinitives are almost identical to the RV. The most striking difference is that accusative object has been generalized by the time that the texts were recorded (20).

(20)-(22) illustrate purpose clauses which regularly have no subordinating marker. The subjects of nonnominal -dyai (IIr *-dhyai) and the datives davoi, aniUe are deleted by coreferenc~ with main clause subject ':'

(20) !caBO, aSai drufJm dyqm zastayo ni him how Truth Lie I-would-place hand down it

dat acc loc 'acc m~€f:Zdyai ()wahya mq()raiS 8~ngkahya ~mavaitim sinqm davoi cast-down your words teaching powerful chasm give inf instr gen ace ace inf dragvasii, a is dva/8ang mazda aniile qstqs- co, (Y 44.14) Liars to them suffering Wise bring hostility and 100 ace aee voe inf ace 'How would I place the Lie in the hand of Truth in order to cast it down by the words of your teaching, in order to put

14

a powerful chasm between the Liars, in Order to bring sufferingand hostility to them, o Wise One?'

The only exception to the accusative object rule of the purposeclauses is one dative:

(21) mqOrwn spdntwn maraeta .. . uxine xraQwe (Vd 4.45)word holy they-speak increase wisdomacc acc sg inf dat'They speak the holy word in order to increase wisdom/

The Infinitive is passivized by promoting patient to subject. In (22)the dative infinitive with passive readingisformedfrom an extend-ed heteroclite stem and has subject deleted by coreference withmain clause object:

(22) aoi mqm staomaine stüiöi yadaP me praise praise in-order-that.

acc inf 2 sg imvmä aparacit saofyantö stavqn (Y 9.2)me future saviors they-praiseacc nom pl nom'Praise me that I may be praised, that future saviors maypraise me/ *

In addition to the purpose clauses, the infinitive complementsa list of subject-equi verbs which are similar to those of the RV(23), (24)7

(23) visaiti dim fräyrärayo (N 19)he-knows him enliven

acc infknows how to enliven him/

As in the RV, wish admits a complement with independent subject(accusative):

7 Av roots in this group are: ae$-, vas- 'wish, desire', - 'be eager', man-'think', vaed- 'know', xa§(y)- 'be able', spam- 'strive', upa-drag- 'finish,succeed*.



14

a powerful chasm between the Liars, in order to bring suffering and hostility to them, 0 Wise One~'

The only exception to the accusative object rule of the purpose clauses is one dative:

(21) '11I48rml. spantml. maraeta . .. uxsne xra{)we (Vd 4.45) word holy they-speak increase wisdom acc acc sg inf dat 'They speak the holy word in order to increase wisdom.'

The infinitive is passivized by promoting patient to subject. In (22) the dative infinitive with passive readingisformedfrom an extended heteroclite stem and has subject deleted by coreference with main clause object:

(22) aoi '11I4m staomaine stuilJi P me praise praise

acc inf 2 sg imv

ya()a in-order-that .

mii aparaci! saosyanto stavqn (Y 9.2) me future saviors they-praise acc nom pI nom 'Praise me that I may be praised, that future saviors may praise me.'

In addition to the purpose clauses, the infinitive complements a list of subject-equi verbs which are similar to those of the RV (23), (24).7

(23) visaiti dim /rayrarayo (N 19) he-knows him enliven

acc inf 'He knows how to enliven him.'

As in the RV, wish admits a complement with independent subject (accusative) :

7 Av roots in this group are: aes-, vas- 'wish, desire', iz- 'be eager' man'think', vaed- 'know', Xa8(Y)- 'be able', Bpar3z- 'strive', upa-drarj- 'finish succeed'. . ,

15

(24) at toi ätrSm ahurä « . . us&maM ... stoithen your fire Lord we-wish be

acc voc inf

rapante ciOrä.avarfidm (Y 34.4)supporter brilliant-helpdat acc'Then we wish, o Lord, that your fire . . . may be of brillianthelp to your supporter/

A morphological difference between BV and Av is that thenominal paradigms which yield the Av infinitives have undergonea certain amount,-of disintegration; often only a few case formsare attested for any action noun. Syntactic differences lie mainlyin the confinement of object case to accusative. Still there is someevidence that Av action nouns are at times indeterminately nomi-nal/infinitival. (25), for example, has a dative verbal abstractfräxSndne 'instruct' which is ambiguously a nominal dative of goalor a passive infinitive.

(25) hyat-cä moi mraoS aSvm, jasö fräxswdne (Y 43.12)this and me you-say Truth you-go instructacc dat acc dat inf/NPnoun: 'This you say to me: "You go to Truth for instruction'Yinfinitive: ' . . .in Order to be instructed/

In contrast, no ambiguity can be detected with verb complements.

2.3 OLD IRISH uses both accusative and dative verbal nounsin subordinate clauses similar to the above examples. Dative verbalnouns (preceded by preposition do 'to'8) are used in purpose clauseswhile accusatives are predicates in complements whose subjectsare either coreferential or independent (and thus overt). The Olrverbal noun appears in the rest of the grammar in many uses,most often äs a simple noun: subject/object of the finite verb,oblique prepositional phrases, etc. This is due to the fact that theverbal noun is completely within the nominal paradigm and nominal

8 Other prepositions may also be used with the verbal noun in purposeclauses (e,g. in with accusative), butthese are not äs numerous; for examplessee Disterheft 150 L



15

(24) a! toi atrm, ahura •.• u.mnahi • .. stoi then your fire Lord we-wish be

acc voc inf

rapante ciOra.ava'Y}ham (Y 34.4) supporter brilliant-help dat acc 'Then we wish, 0 Lord, that your fire ... may be of brilliant help to your supporter.' '

A morphological difference between RV and Av is that the nominal paradigms which yield the Av infinitives have undergone a certain amount(9f disintegration; often only a few case forms are attested for any action noun. Syntactic differences lie mainly in the confinement of object case to accusative. Still there is some evidence that Av action nouns are ap.times indeterminately nominal/infinitival. (25), for example, has a dative verbal abstract /raun'fJne 'instruct' which is ambiguously a nominal dative of goal or a passive infinitive.

(25) hyal-Cii moi mraos aAam Jaso fraun'fJne (Y 43.12) this and me you-say Truth you-go instruct acc dat acc dat infjNP noun: 'This you say to me : "You go to Truth for instruction": infinitive: ' ... in order to be instructed.'

In contrast, no ambiguity can be detected with verb complements.

2.3 OLD IRISH uses both accusative and dative verbal nouns in subordinate clauses similar to the above examples. Dative verbal nouns (preceded by preposition do 'to'S) are used in purpose clauses while accusatives are predicates in complements whose subjects are either coreferential or independent (and thus overt). The OIr verbal noun appears in the rest of the grammar in many uses, most often asa simple noun: subject/object of the finite verb, oblique prepositional phrases, etc. This is due to the fact that the verbal noun is completely within the nominal paradigm and nominal

8 Other pre:positions ma.y a.lso be used with the verbal noun in purpose clauses (e.g. IN with accusative), but these are not as numerous; for examples see Disterheft 150 f.

16

versus infinitival analysis is only determined syntactically. If afüll noun, object is genitive; if pronominal, it is a proclitic pos-sessive pronoun.

The dative verbal noun in purpose clauses has no subordinatorbut is simply placed to the right of the main clause. In (26) subjectis deleted by coreference with a main clause NP and object isa proclitic pronoun suffixed to the preposition da.

(26) dandersaig dia dia m-brith huan chadraighe-stirred-him God for-their carry from-thecity

nom vn-dat dat(Ml 66 c 14)

'God stirred him in Order to carry them from the city/

Independent subject takes the form of agental do plus noun (27)or pronoun (28).9

(27) do atrob do dia and (Wb 21 c 7a)to dwell to Godin-it

vn-dat dat'that God may dwell in it'

(28) du buith dait and fein secech talmain (Ml 84 a 2)to be to-you in-it itself beyond-everjMand

vn-dat'that you should be in it itself beyond every land"

If the verbal noun is intransitive, a proclitic possessive pronoun isoptionally substituted for the agental do phrase ; in such examples,the possessive pronoun subject is, like the possessive object pronounin (26), suffixed to the do which precedes the verbal noun:

(29) dum fortacht huait- siu (Ml 88 a 6)to-my aid by-you PT

vn-dat emph'in order that I be,aided by you'

9 This do phrase would be difficult to consider a true subject since it isalso used äs agent in passive clauses and in nominal constructions.



16

versus infinitival analysis is only determiued syntactically. If a full noun, object is genitive; if pronominal, it is a proclitic possessive pronoun.

The dative verbal noun in purpose clauses has no subordinator but is simply placed to the right of the main clause. In (26) subject is deleted by coreference with a main clause NP and object is a proclitic pronoun suffixed to the preposition do.

(26) dandersaig dia dia m-brith huan chadraig he-stirred-him God for-their carry from-the city

nom vn-dat dat (M166 c 14)

'God stirred him in order to carry them from the city:

Independent subject takes the form of agental do plus noun (27) or pronoun (28).9

(27) do atrob do dia and (Wb 21 c 7a) to dwell to God in-it

vn-dat dat 'that God may dwell in it'

(28) du buith dait and fein secech talmain (M184 a 2) to be to-you in-it itself beyond-every land

vn-dat 'that you should be in it itself beyond every land'

If the verbal noun is intransitive, a proclitic possessive pronoun is optionally substituted for the agental do phrase; in such examples, the possessive pronoun subject is, like the possessive object pronoun in (26), suffixed to·the do which precedes the verbal noun:

(29) dum fortacht huait- siu (MI 88 a 6) to-my aid by-you PT

. vn-dat emph 'in order that I b~ aided by you'

9 This do phrase would be difficult to consider a true subject since it is also used as agent in passive clauses and in noininal constructions.

17

Complements to subject-equi verbs10 take the form of accusativeverbal nouns (30), (31). Object marking is the same s in purposeclauses: genitive noun object (30) or proclitic possessive pro-noun (31).

30) conicimm digail 7 cosc neich (Wb 20 d 6)I-am-able punish and reprimand anyone

vn-acc vn-acc genΊ am able to punish and reprimand anyone/

(31) α n-ru-n-aninraccaigestar dia a ditin- $ora(Ml62b21)wben he-refused God his protect PT

,γ nom vn-aec emph'when God refused to protect him'

Unlike the similar Ilr clauses where coreference triggers man-datory subject deletion, it is optional with active but never appliedto passive verbal nouns (cf. also (29)):

(32) air ni tormenatar- som etir afor NEG they-thought PT at-aUtheir

emphn-ditin 7 an icc (Ml 106 d 11)protect and their savevn-acc vn-acc'for they did not at all think that they would be protectedand saved*

That a n- 'their' is indeed the subject of a passive verbal noun andnot the object of an active one is demonstrated by the fact thatthe verbal noun ditin paraphrases the Latin protecti sunt 'theywere protected' in the passage which the Olr glosses: contra ergoomnem spem Dei auxilio protecti sunt (Ps 86) 'therefore againstevery hope that they would be protected by the help of God'.

Complements to say, declare, think, decide, believe, desire mayhave independent subject in the form of the do agent phrase:

10 These inolude: do-moinethar 'thinks', midithir 'decides', ad-cobra 'desiree',con-icc *is able', dligid *dêrves', ro-tecc diriug 'succeeds', ro-laimethar *dares%intinnecana 'begins'.



17

Complements to subject-equi verbs10 take the form of accusative verbal noup.s (30), (31). Object marking is the same as in purpose clauses: genitive noun object (30) or proclitic possessive pronoun (31).

30) conicimm digail 7 C08C neich (Wb 20 d 6) I-80m-able punish and reprimand anyone

vn-acc vn-acc gen 'I am able to punish and reprimand anyone.'

(31) a n-ru-n-aninraccaigestar dia a ditin- 80m (M162 b 21) when he-refused God his protect PT

.x nom vn-acc emph 'when God refused to protect him'

Unlike the similar IIr clauses· where coreference triggers mandatory subject deletion, it is optional With active but never applied to passive verbal nouns (cf. also (29» :

(32) air ni tormenatar- 80m etir a for NEG they-thought PT at-all their

emph n-ditin 7 an icc (M! 106 d 11) protect and their save vn-acc vn-acc 'for they did not at all think that they would be protected and saved'

That a n- 'their' is indeed the subject of a passive verbal noun and not the object of an active one is demonstrated by the fact that the verbal noun ditin paraphrases the Latin protecti 8unt 'they were protected' in the passage which the OIr glosses: contra ergo omnem 8pem Dei auxilio protecti 8unt (Ps 86) 'therefore aga.inst every hope that they would be protected by the help of God'.

Complements to 8ay, declare, think, decide, believe, de8ire may have independent subject in the form of the do agent phrase:

10 These include: do-moinethar 'thinks', midithir 'decides', ad-cobra 'desires', con-icc 'is able', dligid'deserves', ro-ucc diriug 'succeeds', ro-Zaimethar 'dares', intinnscana 'begins'.

18

(33) .. .is airi asbeir- som a epert doib (Ml 31 b 17)it-is on-account- he-says PT itssay to-them

of-it emph vn-acc'It is on account of it he says that they eay it/

(34) do futhractar for n-imdibe- ei (Wb 20 c 23)they-desire your circumcise PT

vn-acc emph'They desire that you be circumcised/

The complements illustrated in (30)—(34) indicate that eventhough the accusative verbal noun is fully a member of a nominalparadigm, its use corresponds to Infinitive constructions in other

languages. Olr, like Ilr, has accusative verb complementswhich are indeterminate: they may be either embedded comple-ments or simple nouns. This happens in many cases when a genitiveis used whose relationship to the verbal noun is unspecified. Forinstance, in (35) the pronoun a n- 'their' appears before the accu-sative ic 'save':

(35) ni ru frescachtar a n-ic (Ml 26 b 25)NEG they-expected their save

vn-acc

In this sentence, a n- is identical to the subject of expect and maybe interpreted äs a possessive modifying a substantival verbalnoun: 'They did not expect their salvation/ Just äs likely, itcould be a subject pronoun coreferent with the main clause subject,following the pattern of (32): 'They did not expect to be saved/This ambivalence is reflectedîn previous translations, such äsStokes' and Strachan's (1901: 51) äs a noun versus Gagnepain's(1963 : 84) äs a passive infinitive. Compare (36) without theambiv-alent pronoun:

(36) conn- gestais huili taidchorso-that they-would-pray all release

nom vn-accaa indoiri ... (Ml 131 d 13)from captivity

datBrought to you by | University of South Carolina Libraries (University of South Carolina School of L)

Authenticated | 172.16.1.226Download Date | 5/31/12 9:18 PM

18

(33) .• . is at" asbeir- som a epert doib (MI31 b 17) it-is on-account- he-says PT its say to-them

of-it emph vn-acc 'It is on account of it he says that they say it:

(34) do futhractar lor n-imdibe- si (Wb 20 c 23) they-desire your circumcise PT

vn-acc emph 'They desire that you be circumcised.'

The complements illustrated in (30)-(34) indicate that even though the accusative verbal noun is fully a member of a nominal paradigm, its use corresponds to infinitive constructions in other IE languages. OIr, like IIr, has accusative verb complements which are indeterminate: they may be either embedded complements or simple nouns. This happens in many cases when a genitive is used whose relationship to the verbal noun is unspecified. For instance, in (35) the pronoun a n- 'their' appe~rs before the accusative ic 'save':

(35) ni ru frescachtar an-ic (MI 26 b 25) NEG they-expected their save

vn-acc

In this sentence, a n- is identical to the subject of expect and may be interpreted as a possessive modifying a substantival verbal noun: 'They did not expect their· salvation:. Just as likely, it could be a subject pronoun coreferent with the main clause subject, following the pattern of (32): 'They did not expect to be saved: This ambivalence is reflected_ in previous translations, such as Stokes' and Strachan's (1901: 51) as a noun versus Gagnepain's (1963: 84) as a passive infinitive. Compare (36) without the ambivalent pronoun:

(36) conn- gestais huili taidchor so-that they-wo~ld-pray all release

nom vn-acc as indoiri . .. (MI 131 d 13) from captivity

dat

19

where the verbal noun may be analyzed äs direct object of 'pray*('so that they all should pray for release from captivity') or äs itseomplement ('.. .to be released from captivity'). Again past trans-lations have been both nominal (Stokes and Strachan 451) andpassive infinitival (Gagnepain 83).

Purpose clauses are not liable to so much indeterminacy äs theabove accusative complements, but double readings do sometimesobtain. In (37) the enclitic genitive pronoun preceding the verbalnoun may be read äs either the subject of a passive infinitival verbalnoun or äs the object of a nominal one.

(37) an- äs torbe do sochudi dia n-icc (Wb 11 c 6)what it-is prpfit to multitude for-their save

nom dat vn-datnoun: 'What is profit to the multitude for saving them?'infinitive: 'What is profit to the multitude that they besaved?'

(38) is analyzable äs either a passive eomplement with nonappli"cation of equi, äs an active eomplement with a proclitic object,or äs a dative of goal with possessive pronoun: .

(38) .. .condarbais frecndarcus du fortachtae dunaibso-that-they- presence yourhelp to-themay-show gentrebaib so dia soirad ... (Ml 101 c 7)tribes PT to-their deliverdat emph vn-datnoun: 'in order that you may show the presence of your helpto these tribes for their deliverance>

infinitive: '.. .in order that they be delivered'; 'in order todeliver them'

2.4 As mentioned in the Introduction, the morphological simi-larity between Ilr and Celtic is preserved nowhere eise in IE andcan only be considered an archaism. We have seen that purposeclauses are formed in the same way: they are introduced by noconjunction and subject is deleted by coreference with a mainclause NP. TJse of independent, overt subject is found only in RVwith no trace in Av; Olr uses a prepositional agent phrase which



19

where the verbal noun may be analyzed as direct object of 'pray' ('so that they all should pray for release from captivity') or as its complement (' ..• to be released from captivity'). Again past translations have been both nominal (Stokes and Strachan 451) and passive infinitival (Gagnepain 83).

Purpose clauses are not liable to so much indeterminacy as the above accusative complements, but double readings do sometimes obtain. In (37) the enclitic genitive pronoun preceding the verbal noun may be read as either the subject of a passive infinitival verbal noun or as the object of a nominal one.

(37) an- as torbe do sochudi dia n-icc (Wb 11 c 6) what it-is prpfit to multitude for-their save

nom dat vn-dat noun: 'What is profit to the multitude for saving them~' infinitive: 'What is profit to the multitude that they be saved~'

(38) is analyzable as either a passive complement with nonappl( cation of equi, as an active complement with a proclitic object, or as a dative of goal with ~ossessive pronoun: .

(38) .. . concIarbais frecndarcus du fortachtae dunaib so-that-they- presence your help to-the may-show gen trebaib so dia soirad . •• (MI.101 c 7) tribes PT to-their deliver dat emph vn-dat noun: 'in order that you may show the presence of your help to these trihes for their deliverance' infinitive: ' ... in order that they be delivered'; 'in order to deliver them'

2.4 As mentioned in the Introduction, the morphological similarity between IIr and Celtic is preserved nowhere else in IE and can only be considered an archaism. We have seen that purpose clauses are formed in the same way: they are introduced by no conjunction and subject is deleted by coreference with a main clause NP. Use of independent, overt subject is found only in RV with no trace in Av; OIr uses a prepositional agent phrase whioh

20

is probably a later Innovation. In RV, all forme of infinitives .maybe used, but datives do predominate. In Av, assignment of infinitiveform according to clause type no longer obtains. In Olr, dativeverbal nouns (always with preposition do) are predicate in non-finite purpose clauses. Object case of a purposive infinitive neednot be the same äs that of finite verbs: UV object is most oftenaccusative, but dative and genitive are found in substantial num-bers. Av, which has leveled infinitive object to accusative, preservesone or two datives. Olr maintains its strict adherence to genitiveobject of verbal nouns. I suggest that on the basis of these languages,late stages of PIE used the infinitive äs predicate in purpose clauseswith the following features:

a. clause has no subordinator;b. subject is optionally deleted by coreference with a main

clause NP;c. object is usually not marked like that of the finite verb, but

is dative or genitive ;d. infinitive has dative case (at least in earlier stages).The discussion of verb complements so far has been limited to

complements of verbs whose subjects control coreference withlower subject and to complements that have an independent,overt subject. Other classes of verb complements (object-equi andsubject-raising, to be discussed in §3) simply do not show syntacticparallele nor do they have indeterminate analyses. Verbs whosesubjects control coreference in Ilr and Celtic are wish, like, know,fhinky decide, be able, succeed, strive, finish, begin, etc. When anobject appears, it is always accusative in Av and most often so inthe RV; Olr adheres to its genitive object (for accusative, see(48) (49) below). While Av does not show a preference for any spe-cific infinitive class or case, RV has a predominance of accusativeshere and Olr limits this clause type to accusative verbal nouns.Due to the shared features of

a. accusative infinitive,b. genitive or accusative object,c. similarity in ränge of main clause verbs

it appears that complement to at least several of the just-mentionedlist of subject-equi verbs may have had sentential objects withinfinitive in late PIE.

The two clause types for which I have suggested a PIE dateshare an additional feature: they are subject to a certain amount



20

is probably a later innovation. In RV, all forms of infinitives .may be used, but datives do predominate. In Av, assignment of infinitive form according to clause type no longer obtains. In OIr, dative verbal nouns (always with preposition do) are predicate in nonfinite purpose clauses. Object case of a purposive infinitive need not be the same as that of finite verbs: RV object is most often accusative, but dative and genitive are found in substantial numbers. Av, which ha.~ leveled infinitive object to accusative, preserves one or two datives. OIr maintains its strict adherence to genitive object of verbal nouns. I suggest that on the basis ofthese languages, late stages of PIE used the infinitive as predicate in purpose clauses with the following features:

a. clause has no subordinator; b. subject is optionally deleted by coreference with a main

clause NP; c. object is usually not marked like that of the finite verb, but

is dative or genitive; d. infinitive has dative case (at least in earlie:r; stages). The discussion of verb complements so far has been limited to

complements of verbs whose subjects control coreference with lower subject and to complements that have an independent, overt subject. Other classes of verb complements (object-equi and subject-raising, to be discussed in §3) simply do not show syntactic parallels nor do they have indeterminate analyses. Verbs whose subjects control coreference in IIr and Celtic are wish, like, know, think, decide, be able, succeed, strive, finish, begin, etc. When an object appears, it is always accusative in A v and most often so in the RV; OIr adheres to its genitive object (for accusative, see (48) (49) below). While Av does not show a preference for any specific infinitive class or case, RV has a predominance of accusatives here and OIr limits this clause type to accusative verbal nouns. Due to the shared features of

a. accusative infinitive, b. genitive or accusative object, c. similarity in range of main clause verbs

it appears that complement to at least several ofthe just-mentioned list of subject-equi verbs may have had sentential objects with infinitive in late PIE.

The two clause types for which I have suggested a PIE date share an additional feature: they are. subject to a certain~mount

21

of indetermiiiacy in each of the languages by virtue of the factthat they are morphologically identical to nouns which occur inthe same slot in the sentence. They are often to be analyzed äseither accusative direct objects, accusatives after motion verbs,or datives of goal. One additional type does seem to be reconstruct-ible for Ilr but not for other branches: complements to wishwith noncoreferent, overt subject. The fact that this type is alsosubject to indeterminacy indicates an early development, butprobably only within the Ilr period. In contrast, purpose clauseswith overt subject (and others discussed in §3 below) are neverindeterminate. In Olr, the peculiar patterning of subject in bothverb complements and purpose clauses lends itself to multipleanalyses.

3. BEYOND INDETERMINACY

In addition to the structtires discussed above, there are comple-ment types in each of these languages that under no circumstancescan be confused with simple NPs. Furthermore, a lack of similarityin their syntactic features indicates that they have developedindependently. Object-equi clauses are one such type: both KV(39) and Av (40) have such structures.

(39) sa Im mamäda mähi karma kartave (II. 22.1)DEM him he-has great work donom acc inspired acc acc inf'He has inspired him to do great work/

(40) v& naecls darfst ite (Y 43.13)you no-one he-compelled goacc pl nom inf'No one compelled you to go/

It is, however, unlikely that this structure developed before theIlr period: the inventory of main clause verbs11 does not match ;unambiguoue datives are used äs infinitives here; Olr does nothave this strategy.

*help'; Av kav- *pr'incite', dar£~ ômpel̂ dare', yam- 'allow*.



21

of indeterminacy in each of the languages by virtue of the fact that they are morphologically identical to nouns which occur in the same slot in the sentence. They are often to be analyzed as either accusative direct objects, accusatives after motion verbs, or datives of goal. One additional type does seem to be reconstructible for IIr but not for other branches: complements to wish with noncoreferent, overt subject. The fact that this type is also subject to indeterminacy indicates an early development, but probably only within the IIr period. In con tra.st , purpose clauses with overt subject (and others discussed in §3 below) are never indeterminate. In OIr, the peculiar patterning of subject in both verb complements and purpose clauses lends itself to multiple analyses.

,/ .\

3. BEYOND INDETERMINACY

In addition to the structures disc'Q&.sed above, there are complement types in each of these languages that under no circumstances can be confused with simple NPs. Furthermore, a lack of similarity in their syntactic features indicates that they have developed independently. Object-equi clauses are one such type: both RV (39) and Av (40) have such structures.

(39) sa i'lh mamiida muhi karma kartave (II. 22.1) DEM him he-has great work do nom acc inspired ace acc inf 'He has inspired him to do great work:

(40) va naecis dar'iJst ite (Y 43.13) you no-one he-compelled go ace pI nom inf 'No one compelled you to go:

It is, however, unlikely that this structure developed before the lIr period: the inventory of main clause verbsll does not match; unambiguous datives are used as infinitives here; OIr does not have this strategy.

11 Verbs whose objects control equi deiet.ion are: RV i· 'clItreat', mand· 'inspire', mah· 'grant', vr- 'choose', Itu· 'call, invite', cit· 'perceive', avo 'help'; Av kav· 'promise', nwtii~ 'command', daxA-, zAa·, 8qlt- 'teach', hak· 'incite', darA· 'compel. dare', yam· 'allow'.

22

Αν raises Infinitive subject to object of man- 'think' and m rav-'say' by marking it s main clause object with accusat ive case :

(41) af Qw mvnghi paourvim mazd yaz mand you I-thought first Wise last

acc acc voc accatoi τηαηαηΗά (Υ 31.8)be mindinf instr4 And with my mind, o Wise One, I have thought you to be thefirst (and) the last/

RV has no strategies which match Av subject-to-object raising.By the time of Classical Sanskrit the infinitive only has one form,-turn, which has become disassociated from the nominal paradigmsince the Vedic period when its d&iive' (-tave), along with otherstems, was used predominantly in the same function. Thus verbcomplements (42) s well s purpose clauses (43) in the later periodregularly have the same distinct infinitive form (examples from(Cardona 1976: 149).

42) kartum icchati devadattahmake he-wants Devadattainf nom'Devadatta wants to make . . /

(43) bhoktum vrajati devadattaheat he-goes Devadattainf nom'Devadatta is going in order to eat/

However, Classical Sanskrit can still use the dative action nounin a phrase equivalent to (43):

(44) bhojan ya vrajati devadattaheat he-goes Devadattadat nom

Use of the same dative thematic action noun had earlier beenadmitted in the RV s an infinitive, s the accusative object in(45) makes clear:



22

Av raises infinitive subject to object of man- 'think' and m rav'say' by marking it as main clause object with accusat ive case:

(41) al Owa mlJnghi paourvim mazda ytizum and you I-thought first Wise last

acc acc voc acc 8toi manafJM (Y 31.8) be mind inf instr 'And with my mind, 0 Wise One, I have thought you to be the first (and) the last.'

RV has no strategies which match Av subject-to-object raising. By the time of Classical Sanskrit the infinitive only has one form, -tum, which has become disassociated from the nominal paradigm since the Vedic period when its dative' (-tave), along with other stems, was used predominantly in the same function. Thus verb cemplements (42) as well as purpose clauses (43) in the later period regularly have the same distinct infinitive form (examples from (Cardona 1976: 149).

42) lcartum icchati devadattalJ,-make he-wants Devadatta in! nom 'Devadatta wants to make .. .'

(43) bholctum vrajati devadattalJ, eat he-goes Devadatta iuf nom 'Devadatta is going in order to eat.'

However, Classical Sanskrit can still use the dative action noun in a phrase equivalent to (43):

(44) bhojanaya vrajati devadattalJ, eat he-goes Devadatta-dat -- nom

Use of the same dative thematic action noun had earlier been admitted in the RV as an infinitive, as the accusative object in (45) -makes clear: _ -

23

(45) abodhi hotä yajathäya devän (V. 1.2)he~has-awakened sacrificer worship gods

nom inf acc'The sacrificer has awakened in order to worship the gods/

-äya had obviously been part of the pattern of Subordination whichallowed infinitival purpose clauses to use any dative action noun.However, once -turn became dominant, -äya then resumed itsexclusive function äs action noun stripped of former infinitivalproperties.

In Olr the subject of the dative verbal noun may be raised toobject ofthink, know, and hear (46). Here the Ipwer subject has notonly accusative case marking but has been moved to matrix objectPosition preceding verbal noun.

(46) ma-ni fessed comdidnad doif NEG it-knew consolation to' *

accthiarmoracht ind uilc (Ml 87 d 4)follow the evilvn-dat gen*if it did not know that consolation follows eviT

Verbal nouns in subject-raising environments (46) differ in oneimportant aspect from the complements to subject-equi verbs(30)—(32): dative case marking precludes Interpretation äs directobject. This, coupled with the strong infinitival characteristics ofsubject-raising, is an important step away from nominal status.The dative verbal noun, äs an unambiguous embedded predicate,appears to be expanding its use during the Olr period: from onlysubject-raising verbs to include those with coreferential subject.A few examples are found in which the dative verbal noun comple-ments subject-equi verbs: be able, desire, love. Note that for thefirst time, object of the verbal noun is a preposed accusative:

(47) .. .ni cumcat aifhirgi ndo denum (Ml 23 a 5)NEG they-are-able repentanceto do

acc vn-dat'They are not able to do repentance/



23

(45) abodhi Mta yajdthaya devdn (V. 1.2) he-has-awakened sacrificer worship gods

nom inf acc 'The sacrificer has awakened in order to worship the gods.'

-aya had obviously been part of the pattern of subordination which allowed infinitival purpose. clauses to use any dative action noun. However, once -tum became dominant, -aya then resumed its exclusive function as action noun stripped of former infinitival properties.

In OIr the subject of the dative verbal noun may be raised to object of think, know, and hear (46). Here the lower subject has not only accusative case marking but has been moved to matrix object

\' position preceding verbal noun..

(46) ma-ni /e8sed wmdidnad do if NEG it-knew consolation to' .

acc thiarmorru'kt ina uite (Ml 87 d 4) follow the evil vn-dat gen 'if it did not know that consolation follows evil'

Verbal nouns in subject-raising environments (46) differ in one important aspect from the complements to subject-equi verbs (30)-(32): dative case marking precludes interpretation as direct object. This, coupled with the strong infinitival characteristics of subject-raising, is an important step away from nominal status. The dative verbal noun, as an unambiguous embedded predicate, appears to be expanding its use during the OIr period: from only subject-raising verbs to include those with coreferential subject. A few examples are found in which the dative verbal noun complements subject-eqw verbs: be able, desire, love. Note that for the first time, object of the verbal noun is a preposed accusative:

(47) .. . ni cumcat aithirgi nao denum (Ml 23 a 5) NEG they-are-able repentance to do

acc vn-dat 'They are not able to do repentance:

24

The dative verbal noun in any type of complement is rare outsideof the Milan glosses. In the glosses on Olr verse, slightly later inthe period than the biblical glosses, restrictions on subject coref-erence loosen and uee of the dative verbal noun spreads to as-beirin the meaning of 'teil, order' (48). Here the embedded subject isdeleted by coreference with the indirect object of 'teil'. Verbalnoun object is, äs in (47), a preposed accusative.

(48) atrubairt Brenaind fria gilla a chochullhe-told Brendan to-the servant his cloak

nom dat accdo chur forro (LH 335.27)to put upon-them

vn-dat'Brendan told his servant to put his cloak upon them/

(48) has replaced an older pattern where, after teil, order, allowfollowed by prepositional phrase, the verbal noun was accusativewith postposed genitive object:

(49) intan asnindet dia 7 forcongair du doinibwhen he-speaks God and he-orders to people

nom datcomattad a firinne ... (Ml 94 b 3)fulfill his righteousness ,vn-acc gen'when God speaks and Orders the people to fulfill his righteous-ness'

(46)—(48), then, show that in the later Olr period a tendencyto avoid indetermiiiacy was starting: dative verbal nouns couldhave subject raised to main clause object (46), could complementsubject-equi verbs (47), or teil (48). RV and Av have developedaway from widespread indeterminacy also with object-equi struc-tures (39), (40) and Av has raising-to-object (41). While RV stillprefers dative infinitives in purpose clauses, other forms can appearin these clause types, e.g. -dhyai (2), (4), -tavai (u) (3), and accu-satives (6) in purpiose clauses, datives (10) in complements. But bythe classical period, Sanskrit has a morphologically separate form,-turn. Av has no trace of the earlier Infinitive case assignment andallows any infinitive form to be used in all clauses.



24

The dative verbal noun in any type of complement is rare outside of the Milan glosses. In the glosses on OIr verse, slightly later in the period than the biblical glosses, restrictions on subject coreference loosen and use of the dative verbal noun spreads to as-heir in the meaning of 'tell, order' (48). Here the embedded subject is deleted by coreference with the indirect object of 'tell'. Verbal noun object is, as in (47), a preposed accusative.

(48) atr'llbairl Brenaind Iria gilla a ehockull he-told Brendan to-the servant his cloak

nom dat acc do en'llr lorro (LH 335.27) to put upon-them

vn-dat 'Brendan told his servant to put his cloak upon them.'

(48) has replaced an older pattern where, after tell, order, allow followed by prepositional phrase, the verbal noun was accusative with postposed genitive object:

(49) intan asnindet dia 7 /orcongair du d6inih when he-speaks God and he-orders to people

nom dat eomallad a I'rinne . .. (M} 94 b 3} fulfill his righteousness VD-acc gen 'when God speaks and orders the people to fulfill his righteous-ness' "

(46)-(48), then, show that in the later OIr period a tendency to avoid 'indeterminacy was starting: dative verbal nouns could have subject raised to main clause object (46), could complement subject-equi verbs' (47), or tell (48). RV and Av have developed away from widespread indeterminacy also with object-equi structures (39), (40) and Av has raising-to-object (41). While RV still prefers dative infinitives in purpose clauses, other forms can appear in these clause types, e.g. -dhyai (2), (4), -tavai (u) (3), and accusatives (6) in purpose clauses, datives (10) in complements. But by the classical period, Sanskrit has a morphologically separate form, -tum. A v has no trace of the earlier infinitive case assignment and allows any infinitive form to be used in all clauses.

25

4. OTHER INDO-EUROPEAN LANGUAGES

Other Indo-Europeanlanguagespresumablywentthrough a stagesimilar to that described above for Ilr and Celtic. By the timethey are attested, however, most have only one Infinitive with"verbal" characteristics. This is, of course, with the exception ofmodern Irish, which has not developed an infinitive äs a separatecategory. Hittite has two forme, -anna and -(u)wanzi, which havelong been separated from the datives that they were formed from(Benveniste 1935b: 30 f., Kammenhuber 1954: 44 f.). -anna wasoriginally assigned to verbs with ablauting sterns, -(u)wanzi tothe others. Germanic uses *-onom (accusative Tastern; cf. Gothic-an), Baltic and Slayic have a *-ti- form and also a supine (*-tum)to express purpose after verbs of motion; cf. the similar Latinsupine in -um. In separate developments, the Latin and Greekinfinitives have entered the verbal System and even acquiredtense and voice marking. Morphological voice distinctions wereapparently not made in Proto-Italic: Latin has active -re (locative5-stem) versus passive -l (dative root noun)/-n (dative s-stem) whileOscan and Umbrian have one form from accusative *-um. Tensemarking was developed in that period by periphrasis of the appro-priately tensed participle with to be (Leumann et al. 1972: 342).Greek on the other hand shows evidence of a more unified morpho-logical development: -ein (or its dialectal reflexes; Attic-Ionic-een < *-esen; cf. KV -sani) was assigned to present thematic stemsvery early. Likewise middle meaning was acquired at an earlydate by -sthai (a disputed formation; see note 2); but in Homer,active forme can still be used where medio-passive is expected(Schwyzer 1953, I: 805). In contrast, the appearance of -nai(dative n-stem), -menai (dative extended heteroclite), and -men(locative extended heteroclite) for athematic infinitives indicatesthat this was not standardized until much later in the individualdialects.

Returning to syntax, parallele to Ilr dative object case in pur-pose clauses are seen in Hittite even though accusative is theStandard case:

(50) nu SAL.MES ukturiya jiaStiyaS leSSuwanziPT women bones collect

nom dat inf



25

4. OTHER INDO·EUROPEAN LANGUAGES

Other Indo-European languages presumably went through a stage similar to that described above for IIr and Celtic. By the time they are attested, however, most have only one infinitive with "verbal" characteristics. This is, of course, with the exception of modern Irish, which has not developed an infinitive as a separate category. Hittite has two forms, -anna and -(u)wanzi, which have long been separated from the datives that they were formed from (Benveniste 1935b: 30 f., Kammenhuber 1954: 44 f.). -anna was originally assigned to verbs with ablauting atems, -( u )wanzi to the others. Germanic uses *-onom (accusative n-stem; cf. Gothic -an), Baltic and Sla~ic have a *-ti- form and also a supine (*-tum) to express purpose ~ter verbs of motion; cf. the similar Latin supine in -um. In separate developments, the Latin and Greek infinitives have entered the verbal system and even acquired tense and voice marking. Morphologioal voice distinctions were apparently not made in Pro.to-Italic: ,Latin has active ore (locative 8-stem) versus passive -i (dative root noun)!-ri (dative 8-stem) while Oscan and Umbrian have one form from accusative *-um. Tense marking was developed in that period by periphrasis of the appropriately tensed participle with to be (Leumann et 0.1. 1972: 342). Greek on the other hand shows evidence of a more unified morphological development: -ein (or its dialectal reflexes; Attic-Ionic -un < *-esen; cf. RV -8ani) was assigned to present thematic stems very early. Likewise middle meaning was acquired at an early date by -8t'hai (a disputed formation; see note 2); but in Homer, active forms can still be' used where medio-passive is expected (Schwyzer 1953, I: 805). ill contrast, the appearance of -nai (dative n-stem), -menai (dative extended heteroclite), and -men (locative extended heteroclite) for athematic infinitives indicates that this was not standardized until much later in the individual dialects.

Returning to syntax, parallels to IIr dative object case in purpose clauses are Been in Hittite even though accusative is the standard case :

(50) nu SAL.MES ukturiya Oaitiyai ld8uwanzi PT women bones collect

nom dat inf

26

pänzi (KUB XXX 15 1-2)they-go

'The women go to the ukturiya to collect bones/Hittite's dative object is undoubtedly inherited from the samesources äs Hr. It is reasonable to assume that the late PIE tran-sitional infinitives could mark their objects with a case that wasneither nominal (genitive) nor wholly verbal (accusative).

The evolution from fewer to more clause types and the increasein verbs which are admitted to these structures are confirmed byother IE languages. Hittite preserves a syntax that reflects theverbal affiliation of -annal(u)wanzi. Here more verb complementtypes are attested. Not only do a large inventory of subject-equiverbs unambiguously take sentential objects,12 but three havestructures whose objects control equi-deletion : tarn- 'allow, let',

l·' Order', fyalzäi- 'calT, äs in

(51) nu namma kiel SA KUR.KUR.T/Jif LÜ.MES TEMIPT moreover this PT lands men message

gen gen genMAffAE DUTU.&T uwauanzi ÜL tarnäi (KUB XIV Ifront Majesty-my come NEG he-allows Rs. 30 f.)'Moreover, he does not let the messengers of these lands comebefore my Majesty/

The infinitive complements three impersonal verbs (ÜL ära- 'it isnot right (to)', UL ki$- *it is not possible (to)', frandäi- and itsSumerogram SI+SÄ- 'be ordained') with which the lower subjectmay either remain in the oblique case :

(52) nu-mu- kau apiya-ya BISKUE UHUgATTI .PTme PT then and Stormgod Hatti Lord-my

enclpiran tiyauwanzi handäittat (KUB V 6 IV 13)forth gp it-was-ordained

inf12 Amongthem are: $arih-,wek·,ilaliya- 4desire', ep- (reflex),zikk-,$ABÄTU

'take, begin', a$$anu, irhäi-, tarup-, zinna- 'be ready, stop, finish', handalliva·t J i _ _ « 4 1 ^ * » _ T C.1 * i i» 1·» 1 . < « . . ' " ' · . ., . -V . _'dare', mat- *endiare', tarfy- 'be able', waqqar- 4fall short (of)', "kars- *fail,neglect', ZI- 'intend', ling- 'be bound by oath', tarkumäi- 'announce',KARÄBU 'promise', memma- 'refuse'. .



26

pii,nzi (KUB XXX 15 I 1-2) they-go

'The women go to the ulcturiya to collect bones: Hittite's dative object is undoubtedly inherited from the same sources as lIr. It is reasonable to assume that the late PIE transitional infinitives could mark their objects with a case that was neither nominal (genitive) nor wholly verbal (accusative).

The evolution from fewer to more clause types and the increase in verbs which are admitted to these structures are confirmed by other IE languages. Hittite preserves a syntax that reflects the verbal affiliation of -anna/{u)wanzi. Here more verb complement types are attested. Not only do a large inventory of subject-equi verbs unambiguously take sentential objects,12 but three have structures whose objects control equi-deletion: tarn- 'allow, let', watarnalJ,lJ,- 'order', lJ,alziii- 'call', as in

, ,

(51) nu namma kiel SA KUR.KUR.PIM LU.M:ES PEMI PT moreover this PT lands men message

gen gen gen MAQAR nUTU.SI uwauanziUL tarniii (KUB XIV I front Majesty-my come. NEG he-allows Rs. 30 f.) 'Moreover, he doesnot let the messengers of these lands come before my Majesty:

The infinitive complements three impersonal v~rbs (UL ara- 'it is not right (to)" UL lciS- 'it is not possible (to)" lJ,andai- and its Sumerogram SI+SA- 'be ordained') with which the lower subject may either remain in the oblique case:

(52) nu-mu-lcan apiya-ya nISKUR URUlJATTI EN.Y A PT me PT then and Stormgod Hatti Lord-my

encl piran tiyauwanzi lJandaittat (KUB V 6 IV 13) forth go it-was-ordained

inf

III Among them are: sanlJ-,wek~,ilaliya- 'desire', ep- (reflex),zikk- SABATU 'take, begin', a8sanu, irhai:-' tarup-, zinna- 'be ready, stop, finish' handalliya'dare', mat- 'endure', tarlJ- 'be able', waqqar- 'fall short (of)" -kars- 'fail neglectt, ZI- 'intend', ling- 'be bound by oath', tarkumlii-' 'announce'; KARABU 'promise', memma- 'refuse' ..

27

'And then it was ordained that I go forth to my Lord, theStormgod of Hatti/

or be raised to subject:

(53) nu T*INGTR.LUM piran tiyanna SIxSÄ- * (KUBV6IV8)PT god forth go it-was-ordained

nom inf'The god was ordained to go forth'.

Other IE languages which are attested much later also exhibit anexpansion in number of clauses with infinitive, Ard 1975 describesa shift from finite to nonfinite complementation in Old English.In the earlier period the number of such complements was smallbut by the later part of that period, most verb types were repre-sented (p. 16 f.).' 'Thus finite complements to verbs like Order,accuse, permit, expect, intend äs in

(54) pa heht he his geferan, paet hio sohton sumne earme pearfan(AElfred, Bede, 388, 10)

'Then he ordered his companions that they seek a certainpoor beggar/

usually antedate an infinitive complement. Ard argued that theModern English constructions which are synchronically derived viaraising cannot be so diachronically. He went on to show that thenonfinite clause type cannot be derived solely from applying equi-NP deletion to a finite structure since the two types occur side byside in early Old English (p. 23). The data suggest rather that theembedded subject is copied äs object of the main clause verb,pronominalized, and subsequently deleted (p. 27). Baising struc-tures with typical NP-infinitive-NP sequences like

(55) John believes Fred to be a plumber.

synchronically can undergo to-be deletion, yielding two derivedobjects NPs ("predicate adjunct construction", p. 32):

(56) John believes Fred a plumber.

But like complements to object-eqüi clauses, NP-NP sequencescannot be diachronically derived from NP-infinitive-NP: they



27

'And then it was ordained that I go forth to my Lord, the Stormgod of Ratti.'

or be raised to subject:

(53) '11/1.£ DINGffi.LUM piran tiyanna SIXSA.-at (KUBV6IV8) PT god forth go it-was-ordained

nom inf 'The god was ordained to go forth'.

Other m languages which are attested much later also exhibit an expansion in number of clauses with infinitive. Ard 1975 describes a shift from finite to nonfinite complementation in Old English. In the earlier period the number of such complements was small but by the later part of that period, most verb types were represented (p. 16 f.):"~Thus finite complements to verbs like order, accuse, permit, expect, intend as in

(54) pa heht he hiB geferan, jJaet hio 80hton sumne earme pear/an (AElfred, Bede, 388, 10)

'Then he ordered his companions that they seek a certain poor beggar.'

usually antedate an infinitive complement. Ard argued that the Modem English constructions which are synchronically derived via raising cannot be so diachronically. Re went on to show that the nonfinite clause type cannot be derived solely from applying equiNP deletion to a finite structure since the two types occur side by side in early Old English (p. 23). The data suggest rather that the embedded subject is copied as object of the main clause verb, pronominalized, and subsequently deleted (p. 27). Raising structures with typical NP-infinitive-NP sequences like

(55) John believes Fred to be a plumber.

synchronically can undergo to-be deletion, yielding two derived objects NPs ("predicate adjunct construction", p. 32) :

(56) John believes Fred a plumber.

But like complements to object-equi clauses, NP-NP sequences cannot be diachronically derived from NP-infinitive-NP: they

28

predate raised constructions and were one of the sources of NP-infinitive-NP complements to eay and ihink. Thus with NP-NPcomplements, it is not the case that one element has been movedfrom the embedded clause to the main clause with the rest of thecomplement remaining in another clause. Rather the entire con-tent of the embedded complement which remains on the surfacehas been moved to the main clause and has become the surfaceobjects of its verb (p. 40). Ard claimed that the NP-infinitive-NPpattern may possibly have arisen under Latin influence, but haddifficulty explaining its spread in later English periods äs due toLatin style. He noted (p. 42) that this pattern became increasinglyeommon after eay, think äs case distinctions were lost and that casemarking may actually have hindered their acceptance in OldEnglish. Raising-to-subject likewise cannot be ascribed to anysingle change that can be described äs raising in the history ofEnglish. Ard stated:

Structures which are derived via Raising-to-Subject in Modern Englishdeveloped diachronically, in the main, from one or both of the followingsources: 1. impersonal verbs with two arguments — an oblique NP anda sentential complement for which the oblique NP became subjectified;2. personal verbs which occurred in structures of the correct syntacticshape

NP V nonfinite V X

but with semantic Interpretation incompatible with a raising analysis,for which the semantic Interpretation changed to allpw a raising analysis(p. 53).

The history of Latin also indicates an expansion of complementstructures: subject-equi verbs take infinitives in the earliest textsbut other types of nonfinite complementation only appear later.Most Classicists agree that accusative subject of the infinitive inLatin and Greek originated in clauses after order (Lat iubeo,Gk keleo) (Schwyzer II: 373 f., Leumann et al. II: 353 f.; allLatin examples are taken by Hahn 1950 from Plautus andTerence):

(57) iube hunc abire (ET 585—6). order him go

imv acc infOrder him to go/



28

predate raised constructions and were one of the sources of NPinfinitive-NP complements to /Jay and think. Thus with NP-NP complements, it is not the case that one element has been moved from the embedded clause to the main clause with the rest of the complement remaining in another clause. Rather the entire content of the embedded complement which remains on the surface has been moved to the main clause and has become the surface objects of its verb (p. 40). Ard claimed that the NP-infinitive-NP pattern may possibly have arisen under Latin influence, but had difficulty explaining its spread in later English periods as due to Latin style. He noted (p. 42) that this pattern became increasingly common after 8ay, think as case distinctions were lost and that case marking may actually have hindered their acceptance in Old English. Raising-to-subject likewise cannot be ascribed to any single change that can be described as raising in the history of English. Ard stated:

Structures which ~e derived via Raising-to-Subject in Modern English developed diachronically, in the main, from one or both of the following sources: 1.. impexsonal verbs with two arguments. -. an oblique NP and a sentential complement for which the oblique NP became subjectified; 2. personal verbs which OCCUlTed in structures of the correct syntactic shape

NP V nonfinite V X

but with semantic interpretation 'incompatible with a raising analysis, for which the semantic interpretation changed to aJ,low a raising analysis (p. 53).

The history of Latin also indiQates an expansion of complement structures: subje~t-equi verbs take infinitives in the earliest texts but other types of nonfinite complementation only appear later. Most Classicists agree that accusative subject ·of the infinitive in Latin and Greek originated in clauses after order (Lat iubeo, Gk keleo) (Schwyzer II: 373 f., Leumann et a1. II: 353 f.; all Latin examples are taken by Hahn 1950 from Plautus and Terence) :

(57) iube hunc abire (HT 585-6) order him go imv acc inf 'Order him to go:

29

(58) tt reksai he Mleue ... Achaious (Iliad 2.11)arm DEM he-ordered Aohaeansinf nom accΉΘ ordered the Achaeans to arm themselves/

Hahn 1950 showed how verbs of perception (= object-equi verbs)very early took a sentential object consisting of a participle modifyingand agreeing with the accusative object noun; thus NP(acc)-participle(acc) obtained in sentences like

(59) vidi et illam et hospitem complexamI-saw and her and guest hug

acc acc ptc-accosculantem (Mil 533—4)

ptc-accΊ saw her and the guest hugging and kissing/

Othertimes the participle is used ambiguously:

(60) meminestin me gravidam? (ET 626)you-remember me pregnant

acc ptc-acc

yielding interpretation s attributive (NP-participle):'Do you remember me being pregnant?'

or predicative (NP-to 6e-partieiple):'Do you remember that I was pregnant?'

Sometimes hear can only have a sentential complement in whichcase a predicative relationship is established between the accusativeNP and its modif ier:

(61) faenus creditum audio (Most 629)interest credit I-hearacc ptc-accΊ hear that it has been credited for the interest/

Hahn then demonstrated how sino 'allow' and patiar 'suffer' areused both with Infinitive:



29

(58) tlWreksai he Mleu.e . .. AchaioUs (Iliad 2.11) arm DEM he-ordered Achaeans inf nom acc 'He ordered the Achaeans to arm themselves.'

Hahn 1950 showed how verbs of perception (= object-equi verbs) very early took a sentential object consisting of a participle modifying and agreeing with the accusative object noun j thus NP(acc)participle(acc) obtained in sentences like

(59) vidi et illam et Mspitem complexam I-saw and her and guest hug

acc acc ptc-acc osculantem (¥iI533-4) kiss .• (

ptc-acc 'I saw her and the guest hugging and kissing.'

Othertimes the participle is used ambiguously:

(60) meminestin me gravidam? (HT 626)

you-remember me pregnant acc ptc-acc

yielrung interpretation as attributive (NP-participle): 'Do you remember me being pregnant~'

or predicative (NP-to be-participle): 'Do you remember that I was pregnant r

Sometimes hear can only have a sentential complement in which case a predicative relationship is established between the accusative NP and its modifier:

(61) faenus creditum audio (Most 629)

interest credit I-hear ace ptc-acc 'I hear that it has been credited for the interest.'

Hahn then demonstrated how sino 'allow' and patiar 'suffer' are used both with infinitive:

30

(62) siguidem potes esse te pati in lepido loci (Poen 696)if you-are- be you suffer in pleaeant place

able inf dat inf loc loc'if you can suffer to be in a pleasant place'

or with the NP-participle construction:

(63) vosne ego patiar cum mendicisyou-PT I I-will-allow with pooracc pl dat plnuptas . .. viris? (Stich 132)married menptc-acc dat pl'So shall I allow you (to be) married to poor men V

Since no examples can be found with present participle or withnonparticipial adjectives, Hahn surmised (p. 121) that in each ofthe examples cited, we have not a participle but a past tense infin-itive (which is formed from past participle plus to be) with copuladeleted. Thus in

(64) abductam illam aegre pati (Merc 251)abducted her scarcely sufferptc-acc acc inf'to scarcely suffer her (to be) abducted'

l

abductam is underlyingly abductam esse with copula (inf esse)deleted, äs is regulär in the rest of Latin grammar. On analogy tosuch constructions with past infinitive, present infinitives, whichare not formed periphrastically, are possible:

(65) sed tu enumquam piscatorem vidisti .. . piscembut you ever fisherman you-saw fish

acc acccepisse? (Rud 987—8)catchinf'But did you ever see a fisherman catch a fish?'

Other Latin clause types also exhibit a gradual spread throughthe ränge of possible structures. Saltarelli 1979 demonstrated that



30

(62) siquidem potu use te pati in lepido loci (Poen 696) jf you-are- be you suffer in pleasant place

able inf dat inf loc loc 'if you can suffer to be in a pleasant place'

or with the NP-participle construction:

(63) VOSM ego patiar cum merulicis you-PT I I-will-allow with poor acc pI dat pI nwptas ... viris? (Stich 132) married men ptc-acc dat pI 'So shall I allow you (to be) married to poor men t'

Since no examples can be found with present participle or with nonparticipial adjectives, Hahn surmised' (p. 121) that in each of the examples cited, we have not a participle but a past tense infinitive (which is formed from past participle plus to be) with copula deleted. Thus in

(64) abductam illam aegre pati (Merc 251) abducted her scarcely suffer ptc-acc acc inf 'to scarcely suffer her (to be) abducted'

abductam is underlyingly abductam use with copula (inf u8e) deleted, as is regular in the rest of Latin' grammar. On analogy to such constructions with past infinitive, present infinitives, which are not formed periphrastically, are possible:

(65) 8ed tu enumquam pi8cat0r..em vidi8ti . .. piscem but you ever fisherman you-saw fish

acc cepisse? (Rud 987~8) catch inf

acc

'But did you ever see a fisherman catch a fish "

Other Latin clause types also exhibit a gradual spread through the range of possible structures. Saltarelli 1979 demonstrated that

31

the following clause types are sequeiitially developed by causativeswith facere 'make*:

a. finite complement with subordinator ut;b. finite complement without ut;c. subject-to-object raising with finite complement;d. subject-to-object raising with active and passive infinitives.

In Romance, clause union competes with the reflexes of facere andis currently the only possibility with Italian fare and Spanishfaire.

The material presented here from other IE languages indicatesa continuing development in each from fewer to more clause types.Like purpose clauses and subject-equi complements, raising andobject-equi structures in Old Bnglish and Latin evolve from mate-rial in the main clause which develops properties of a predicate.All these clausal innovations add further evidence to Ard's claimthat diachronic processes cannot be assumed to recapitulate asynchronic description. While the rise of the four clause typesinvolves a reanalysis of main clause. surface strings, the processesare different. Purpose clauses and subject-equi complements devel-oped at a time when the infinitive was still an action noun andpart of nominal paradigms. In its capacity äs the object of a finiteverb or äs a dative adding material to the sentence, the actionnoun moved into clausal status äs object complement or äs a pur-pose clause, respectively, upon developing subject coreference. Onthe other hand, we find object-equi and raising developing at a timewhen the infinitive was completely separated from nominal para-digms and had acquired more properties of a predicate. In Latinwe see NP-accusative" participle sequences reanalyzed äs NP-infinitive after object-equi verbs. In Old English, raising to objectevolved from NP-NP sequences being reanalyzed äs NP-infinitive-NP. Even though the Old English and the Latin studies are lan-guage-specific descriptions, the mechanisms involved are generaltendencies that can probably be observed for other languagesdeveloping such structures.

5. CONCLUSIONS

In comparing the syntax of the verb complements and purposeclauses in KV, Av and Olr, it is clear that the Olr verbal noun'scompletely nominal status is morphologically closer to PIE than



31

the following clause types are sequentially developed by causatives with facere 'make':

a. finite complement with subordinator ut; b. finite complement without ut; c. subject-to-object raising with finite complement; d. subject-to-object raising with active and passive infinitives.

In Romance, clause union competes with the reflexes of facere and is currently the only possibility with Italian fare and Spanish faire.

The material presented here from other IE languages indicates a continuing development in each from fewer to more clause types. Like purpose clauses and subject-equi complements, raising and object-equi structures in Old English and Latin evolve from material in the main clause which develops properties of a predicate. All these clausalflmovations add further evidence to Ard's claim that diachronic processes cannot be assumed to recapitulate a

_ synchronic description. While the rise of the four clause types involves a reanalysis of main clause. surface strings, the processes are different. Purpose clauses and subject-equi complements developed at a time when the infinitive was still an action noun and part of nominal paradigms. In its capacity as the object of a finite verb or as a dative adding material to the sentence, the action noun moved into clausal status as object complement or as a purpose clause, respectively, upon developing subject coreference. On the other hand, we find object-equi and raising developing at a time when the infinitive was completely separated from nominal paradigms and had acquired more properties of a predicate. In Latin we see NP-accusative' participle sequences reanalyzed as NPinfinitive after object-equi verbs. In Old English, raising to object evolved from NP-NP sequences being reanalyzed as NP-infinitiveNP. Even though the Old English and the Latin studies are language-specific descriptions, the mechanisms involved are general tendencies that can probably be observed for other languages developing such structures.

5, CONCLUSIONS

In comparing the syntax of the verb complements and purpose clauses in RV, Av and OIr, it is clear that the OIr verbal noun's completely nominal status is morphologically closer to PIE than

32

tho other languages which have some infinitives not part of nounparadigme. Olr syntax, however, contains both highly archaicfeatures and some innovations. Archaisms are most striking withsubject of accusative complements: equi-deletion is optional andovert subject is either an agent phrase (with do) or a genitivewhose exact relationskip to the verbal noun is unspecified. Admit-tanee of dative verbal noun complements represents a developmentwithin the later Olr period: it starte to appear in raising-to-objectclauses while subject-equi complements have accusative object.BV has developed further, both morphologically (having nonnom-inal infinitives) and syntactically (object being most offcen dativeor accusative; less restrictions on coreferent and independent sub-ject). Av has gone the furthest by having no restrictions on assign-ment of infinitive form and by standardization of accusative object.All three languages are subject to indeterminacy, but only in theolder clause types.

On the basis of this evidence, I concluäe that these features doindicate that PIE used Subordination with nominalizations. Theinfinitive was, however, not morphologically separate äs it is inmost of the historically attested daughter languages. Syntacticfeatures alone decided when the action noun should be interpretedäs a noun or äs an infinitive. A cpmparison of Ilr and Olr yieldsfeatures so similar that they must have developed during theperiod of unity. Equi-deletion was a PIE development with verbswhose subjects control coreference in purpose clauses.13 From Olr(32) and scattered KV evidence14 equi-deletion may have beenoptional in PIE. Latin's admittance of coreferent infinitive sub-ject after wish may also be a holdover from optional applicationin PIE. Differences in object-equi constructions and the absenceof raising in RV mitigates against their reconstruction for PIE.Genitive was probably the majority object case while dative — tojudge from Ilr and Hittite — was probably a transition case insubordinate clauses. Accusative was an independent development

13 My evidence from Ilr, Celtic, and Hittite (Disterheft 182—4) contra·dicts Müler's 1974: 230 argument from Latin and Greek (and the inferencethat this is true for all IE) that equi NP deletion developed independentlyin these languages. *~

14 RV admits a coreferent subject pronoun in the followig purpose clauses:tcvrl$ani VI. 37.7, itfaye I. 113.5, 6, ütaye VIII. 21.9, pltaye I. 16.3, pra-yaiX. 104.3, taräya VIII. 96.1, madäya IX. 109.20.



32

the other languages which have some infinitives not part of noun paradigms. OIr syntax, however, contains both highly archaic features and some innovations. Archaisms are most striking with subjeot of acousative complements: equi-deletion is optional and overt subject is either an agent phrase (with do) or a genitive whose exact relationship to the verbal noun is unspecified. Admittanoe of dative verbal noun complements represents a development within the later OIr period: it starts to appear in raising-to-object clauses while subject-equi complements have aocusative object. RV has developed further, both morphologically (having nonnominal infinitives) and syntactically (object being most often dative or accusative; less restrictions on coreferent and independent subject). Av has gone the furthest by having no restrictions on assignment of infinitive form and by standardization of aocusative object. All three languages are subject to indeterminacy, but only in the older clause types.

On the basis of this evidence, I concluae that these features do indicate that PIE used subordination with nominalizations. The infinitive was, however, not morphologically separate as it is in most of the historically attested daughter languages. Syntactic features alone decided when the action noun should be interpreted as a noun or as an infinitive. A comparison of Ilr and OIr yields features so similar that they must have developed during the period of unity. Equi-deletion was a PIE development with verbs whose subjects control coreference in purpose clauses.lS From OIr , (32) and scattered RV evidence14 equi-deletion may have been optiOIl,al in PIE. Latin's admittance of coreferent infinitive subject after wish may also be a holdover from optional application in PIE. Differences in object-equi constructions and the absence of raising in RV mitigates against their reconstruction for PIE. Genitive was probably the majority object case while dative - to judge from Ilr and Hittite - was probably a transition case in subordinate clauses. Accusative was an independent development

13 My evidence from TIr, Celtic, and Hittite (Disterheft 182-4) contra; dicts M;iller's 1974: 230 argument from Latin and Greek (and the inference ~hat this is true for all IE) that equi NP deletion developed independently m these languages. ~

14 RVadmits a. coreferent subject pronoun in the folIo wig purpose clauses: tari~ani VI. 37.7, ~lay6 I. 113.5,6, utaY6 VIII. 21.9, pitay6 I. 16.3, pra-yai X. 104.3, tariiya VIII. 96.1, mMiiya IX. 109.20.

33

in each subgroup (äs Olr shows) äs a result of the development ofmore "verbal" infinitive syntax.

All IE languages suryeyed here uniformly display a trendtoward an expansion in types of clauses that admit infinitive andin the number of verbs that use infinitive complements withineach of the semantic classes. After incipient purpose clause andsubject-equi complements of late PIE have become established äsproductive complements, object-equi clauses start to develop äsearly äs Hittite and are already proliferating in Av and RV. Olrdoes not use object-equi per se, but a semantically parallel typewhich deletes a verbal noun's subject when coreferent with theobject of a prepositional phrase (literally: He Orders to the people(= prep phretse) to do this; for examples see Disterheft 145 f.).By the time of Greek and Latin this type is strongly attested.Raising-to-object £- not found in RV or Hittite — is starting inAv and is quite developed in Olr. Raising-to-subject is a distinctlylate type. Its presence has been noted for three verbs in Hittite(52), (53), but it does not appear in .other languages discussed hereuntil Latin, Greek, and Olr (Disterheft 152—5). We have seeninfinitive complementation start in the IE period with subject-equiand purpose clauses, then spread independently in the subgroupsfrom object-equi to raising-to-object then raising-to-subject, alläs the result of reanalysis of surface strings within the mainclause.15

In this discussion I have dealt only with a portion of the IElanguage area; most of the languages omitted are attested laterand are arguably not äs valuable for reconstruction. Furtherinvestigations into the development of structures in other languagegroups which earlier were infinitiveless will ascertain how muchof the successive spread of clause types can be attributed to cross-linguistic tendencies.

DOROTHY DISTERHEFTDepartment of EnglishUniversity of South CarolinaCOLUMBIA, SO 29208USA

15 The chronology pf other constructions not dealt with here (e.g. impera-tive infinitives, predicate infinitives, noun and adjective complements) arodiscussed in Disterheft 181—92.



33

in each subgroup (as Olr shows) as a result of the development of more "verbal" infinitive syntax.

All IE languages surveyed here uniformly display a trend toward an expansion in types of clauses that admit infinitive and in the number of verbs that use . infinitive complements within each of the semantic classes. After incipient purpose clause and subject-equi comple~ents of late PIE have become established as productive complements, object-equi clauses start to develop as early as Hittite and are already proliferating inAv and RV. Olr does not use object-equi per se, but a semantically parallel type which deletes a verbal noun's subject when coreferent with the object of a prepositional phrase (literally: He orders to the people (= prep phrase) to do this; for examples see Disterheft 145 f.). By the time of Greek and Latin this type is strongly attested. Raising-to-object\·e.- not found inRV or Hittite - is starting in Av and is quite developed in Olr. Raising-to-subject is a distinctly late type. Its presence haS been noted for three verbs in Hittite (52), (53), but it does not appear in other languages discussed here until I .. atin, Greek, and Olr (Disterheft 152-5). We have seen infinitive complementation start in the IE period with subject-equi and purpose clauses, then spread independently in the subgroups from object-equi to raising-to-object then raising-to-subject, all as the result of reanalysis of surface strings· within the mam clause.I5

In this discussion I have dealt only with a portion· of the IE 1l1nguage area; most of the languages omitted are attested later and are arguably not as valuable for reconstruction. Further investigations into the development of structures in other language groups which earlier were infinitiveless will ascertain how much of the successive spread of clause types can be attributed to crosslinguistic tendencies.

DOROTHY DISTERHEFT Department of English University of South Carolina COLUMBIA, SC 29208 USA

IS The chronology of other constructions not dealt with here (e.g. imperative infinitives, predicate infInitives, noun and adjective complements) aro discussed in Disterheft 181-92.

34

REFERENCES

ABD, WILLIAM JOSH (1975). Raising and Word Order in Diachronie Syntax.UCLA dissertation reprodueed and dietributed by Indiana LinguisticsClub, 1977.

BABTHOLOMAE, CHBISTIAN (1890). "Das griechische Infinitivsuffix -cr#at",Rheinisches Museum 45: 151—3.

BENVENISTE, EMILE (1935a). Lee injinitifs avestiques. Paris: Adrien Maison-neuve.— (193öb) Originee de la formation des noms en indo-europeen. Paris:Adrien Maisonneuve.

CABDONA, GEOBGE (1976). "Some Features of Päninian Derivations",History of linguistic thought and contemporary linguistics (ed. H. Parret),Berlin: de Gruyter, 137—58.

DrtxoN, MYLES (1955). "On the Syntax of the Irish Verb", TPhS 1955:104-16.

DISTEBHEFT, DOBOTHY (1980). The Syntactic Development of the Infinitivein Indo-European. Columbus, Ohio: Slavica.

GAGNEPAIN, JEAN (1963). La syntaxe du nom verbal dans les langues ceUiques.I. Irlandais. Paris: Klineksieek.

GBASSMANN, HEBMANN (1873). Wörterbuch zum Rig-Veda. [Reprinted byWissenschaftliche Buchgesellschaft, Darmstadt, 1964.]

HAHN, E. ADELAIDE (1950). "The Infinitive with Subject Accusative",81: 117-29.

KAMMENBTTBEB, ANNELIES (1954). Studien zum hethitischen Infinitivsystem.I. Die Morphologie der Infinitive auf -anna (-atar) und -(u)wanzi (-uwar),MIO 2: 44-77.

KLEIN, JABED (1978). The Partide u in the Rigveda: A Synchronic andDiachronie Study (Ergänzungshefbe zur Zeitschrift- für vergleichendeSprachforschung 27). Göttingen: Vandenhoeck & Ruprecht.

LEUMANN, MANU, J. B. HOFMANN, ANTON SZANT^B (1972). LateinischeGrammatik, Munich: Beck.

MUXEB, D. GABY (1974). "On the History of Infinitive Complementation inLatin and Greek", JIES 2: 223—46.

RENOÜ, Louis (1954). Le passage du nom d'action a Vinfinitif dans le Rigveda,Sprachgeschichte und Wortbedeutung (Fs. Albert Debrunner).* Bern:Francke, 383—7.

SALTABEIXI, MABIO (1979). "Syntactic Diffusion". Paper read at the FourthInternational Conference on Historical Linguistics, Stanford.

SCHWYZEB, EDUABD (1953). Griechische Grammatik.2 Munich: Beck.STOKES, WHTELEY and JOHN STBACHAN (1901). Thesaurus Palaeohibernicue I.

[Reprinted by the Dublin Institute for Advanced Studies, 1975.]THUBNEYSEN, RUDOLPH (1908). "Altindisch etavai", Melanges de linguistique

offerts a M. Ferdinand de Saussure. Paris: Champion, 225—7.— (1946) A grammar of Old Irish. Dublin: Institute for Advanced Studies.



34

REFERENCES

ARD, WILLIAM .Tosu: (1976). Raising and Word Order in Diachronic Syntax. UOLA dissertation reproduced and distributed by Indiana Linguistics Club, 1977. .

BARTHOLOMAE, C:a:RISTJAN (1890). "Das griechische Infinitivsuffix -afia.£", Rheini8che8 MU8eurn 45: 151-3.

BENVENISTE, EMILE (1935a). Lu in/initi/s avutiquu. Paris: Adrian Maisonneuve. - (1935b) Origines de la formation des nOmIJ en indo-europeen. Paris: Adrien Maisonneuve.

CARDONA, GEORGE (1976). "Some Features of Pii.r.llnian Derivations", History o/linguistic thought and contemporary linguistics (ed. H. Parret), Berlin: de Gruyter, 137-58.

DILLON, MYLES (1955). "On the Syntax of the Irish Verb", TPhS 1955: 104-16.

DISTER:a:EFT, DOROTHY (1980). The Syntactic Development of the Infinitive in Indo-European. Columbus, Ohio: Slavica.

GAGNEPAlN, JEAN (1963). La syntaxe du nom verbal danalulangUe8 celtiqUeB. 1. Irlandais. Paris: Klinoksieok.

GRASSMANN, HERMANN (1873). WOrterbuch zum Rig-Veda. [Reprinted by Wissensohaftliche Buohgesellschaft, Darmstadt, 1964. J

HAHN, E. ADELAIDE (1950). "The Infinitive with Subjeot Acousative", TAPA 81: 117-29.

KAMMENHUBER, ANNELIES (1954). Studien zum hethitisohen Infinitivsystem. 1. Die Morphologie der Infinitive auf -anna (-atar) und -(u)wanzi (-uwar) , MIO 2: 44-77.

KLEIN, JARED (1978). The Particle u in the Rigveda: A Synchronic and Diachronic Study (Erganzungshefte zur Zeitsohrift· fiir vergleiohende Spraohforschung 27). Gottingen: Vandenhoeok & Rupreoht.

LEUMANN, MANu, J. B. HOFMANN, ANTON SZANTYR (1972). Lateinische Grammatik. Munioh: Beok.

MILLER, D. GARY (1974). "On the History of Irifinitive Complementation in Latin and Greek", JIES 2: 223-46.

RENOU, LOUIS (1954). Le pas8age du nom d'aceion a l'infinitij danale Rigveda, Sprachgeschichte und Wortbedeutung (Fs. Albert Debrunner).· Bern: Francke, 383~7. . .

SALTARELLI, MARIo (1979). '~Syntaotio Diffusion". Paper read at the Fourth International Conferenoe on Historioal Linguistics, Stanford.

SCRWYZER, EDUARD (1953). Griechische Grammatik.2 Munich: Beck. STOKES, WHITLEY and JOHN STRACHAN (1901). TheBaurU8 Palaeohibernicu8 I.

[Reprinted by the Dublin Institute for Advanoed Studies, 1975.] TRURNEYSEN, RuooLl'lI (1908). "Altindisoh.etavai", Melanges de linguistique

ojfert8 a M. Ferdinand de SaU88ure. Paris: Champion, 225-7. - (946) A grammar of Old Irish. Dublin: Institute for Advanoed Studies.

Documents

Remarks on the History of the Indo-European Infinitive