Search

This article presents and discusses evidence that genitive and dative objects regularly become nominative in Ancient Greek passives of monotransitives and ditransitives. This is a typologically and theoretically significant state of affairs for two reasons. (i) As is well known, nonaccusative objects are, in many languages, not allowed to enter into Case alternations, a fact that has been accounted for in the government-binding/principles-and-parameters literature on the basis of the assumption that nonaccusative objects—prototypically datives—bear inherent, lexical , or quirky Case. By this reasoning, Ancient Greek genitives and datives must be concluded to have structural Case. (ii) Even in languages where dative-nominative (DAT-NOM) alternations do obtain, they are often limited to ditransitives, a fact that can been taken to suggest that dative qualifies as structural Case only in ditransitives. A language like Ancient Greek, which allows genitive and dative objects to become nominative in all passives (monotransitives and ditransitives), shows that it is, in principle, possible to have a linguistic system where genitive and dative qualify as structural Cases in both monotransitives and ditransitives. Case theories must be designed in such a way as to allow for this option. We argue for an analysis of Case alternations that combines the view that alternating datives and genitives enter the formal operation Agree with a morphological case approach to the distribution of overt case morphology. We furthermore compare Ancient Greek DAT-NOM and genitive-nominative (GEN-NOM) alternations in passives to Icelandic DAT-NOM and GEN-NOM alternations in middles, pointing to a number of interesting differences in the two types of alternations that depend on (i) the types of nonaccusative arguments entering Agree, (ii) the verbal head (Voice or v) entering Agree with nonaccusative objects, and (iii) the rules of dependent case assignment in connection to the role of nominative in the two languages.

As the Dyirbal language of Australia nears extinction, its diminishing social function is accompanied by reduction within the linguistic system. If we place semi-speakers on a continuum according to the degree to which their language has been simplified, it is possible to trace changes in the grammar of the dying language. Focusing on morphological and syntactic ergativity demonstrates that decay in terminal Dyirbal involves systematic variation along the continuum. Reduction occurs on various levels, affecting syntactic operations and morphology. However, the data suggest that important formal and functional differences exist between dying Dyirbal and a typical pidgin.* This paper investigates structural change in the Dyirbal language of Australia as it approaches extinction. Radical changes have occurred in the dying language, affecting both morphological and syntactic ergativity. Such reduction in the linguistic system, however, is far from a patchwork of ad-hoc errors. The patterned nature of reduction in language death is apparent in the systematic collapse of ergative allomorphs and in the gradual weakening of the S/O pivot operation (cf. Dixon 1979). In ? 1, a sociolinguistic sketch of dying Dyirbal is provided. In ?2, 1 deal with the semi-speakers of the language. In ?3, allomorphic reduction of the ergative affix is described; in ?4, I show changes in syntactic ergativity. In ?5, I present a comparison of changes in ergative case-marking and in S/O pivot operation along the semi-speaker continuum. Finally, ?6 investigates the light that these findings shed on general linguistic debate. SOCIOLINGUISTICS 1.1. THE SETTING. Dyirbal was originally spoken over more than 8,000 square kilometers in the rain forest of north-east Queensland; some six contiguous tribes spoke what can be regarded as dialects of a single language. The population of each tribe was approximately 500. Because of the impact of European civilization (white expansion began in the 1860's), the traditional tribal unit disintegrated: territory was invaded, religious sites were desecrated, and the physical environment was deeply bruised by the impact of clearing the land for pasture and 'progress'. Some Aboriginal people were transported to * I am most grateful to R. M. W. Dixon, Nancy Dorian, and Jane Hill for helpful suggestions on this paper. I would also like to express my gratitude to George Watson and to the people of the Jambun community for their warm friendship and willingness to share their knowledge of Dyirbal with me. The following abbreviations are used in this paper: A, transitive subject; ACC, accusative; ANTIPASS, antipassive derivational affix; ASP, aspectual; DAr, dative; DU, dual; ERG, ergative; INTR, intransitive; LOC, locative; NOM, nominative; NONFUT, non-future; NP, noun phrase; 0, transitive object; PL, plural; PURP, purposive; REDUP, reduplicated; REL, relative clause marker; S, intransitive subject; TD, Traditional Dyirbal; TR, transitive; YD, Young Dyirbal. Individual speakers are also referred to by initials; e.g., LN, EM. The following orthographical differences to Dixon 1972 should be noted: j = Dixon's /l/ (lamino-palatal stop), r = // (retroflex), rr = Irl (trill), ny = /j/ (laminopalatal nasal).

The essays collected in this book result from recent work in trace theory, the starting point being Chomsky's proposals to permit syntax to over-generate profusely and then to provide most of the significant reduction in the form of a system of filters and constraints on logical form. They all address various aspects and implications of his article, "On Binding, "which is reprinted here. Equally important, the essays demonstrate the applicability of Chomsky's proposals to a number of different languages.Heny writes that "editorial work on the volume brought me to a realization of the far-reaching implications of the approach to linguistic analysis underlying Chomsky's "On Binding. "It constitutes a radical break with his previous work, of a very exciting and promising kind." Heny's Introduction goes on to describe the shift in Chomsky's position over the last ten years in lucid and straightforward fashion, identifying the striking new properties of this framework and making the collection generally more accessible to a wider audience. Contents: Introduction; "On Binding, "Noam Chomsky; "Finiteness and Boundedness in Turkish, "Lelan George and Jaklin Kornfilt; "Nominative Marking in Italian Infinitives and the Nominative Island Constraint, "Luigi Rizzi; "Empty Subjects, Case and Agreement and the Grammar of Dutch, "Eric J. Reuland; "Binding, Quantifiers, Clitics and Control," Richard S. Kayne; "Government and Relativisation in Celtic," Stephen Harlow; "Germanic Word Order and the Format of Surface Filters," Joan Maling and Annie Zaenen; "Ouechua Word Structure," Pieter Muysken; "Index."Frank Heny is Professor of Linguistics and the Philosophy of Language, and Head of the Institute of General Linguistics at the University of Groningen, Holland. Distributed in North America for Croom Helm Publishers Ltd.

The development of the original diphthongs in Hittite has seemed to be a peculiarly difficult problem; but evidence has gradually accumulated' until something like a system has emerged. We may now say that all the original short diphthongs appear in Hittite as monophthongs, whereas the original long diphthongs retain their diphthongal character. It is not always possible to distinguish with certainty between the several diphthongs; but it will promote clearness if I treat them separately, as far as possible. Primitive Indo-Hittite"l ai (ai) appears in the nominative plural neuter of the pronominal declension (kue 'quae', ke 'haec', ape 'illa', i, 8e 'ea', te ne 'et ea', udne 'countries') and in the medial ending -ti, which corresponds with Greek -rae and Sanskrit -te (op. cit. 144). Probably another example is afforded by the adjective idalu' 'bad', which I would connect with Gk. aZhaXos 'smoky flame', ailahX6M 'soil with soot', passive 'be laid waste by fire', Old Saxon Tdal, idil, Old English idel, Old High German ital 'idle'. The variation in orthography (e:i) will be discussed below.

Every Serbo-Croatian verb form' consists of the constituents listed in Table 1.2 While prefixes are optional, every verb form has a stem and at least one inflectional element. The stem consists of at least a root.3 Every verb paradigm has a thematic vowel in some or all of its forms, though some forms of some paradigms are without this. The remaining stem constituents are optional. The constituents always occur in the order given in Table 1. A given Serbo-Croatian verb has the following sets of forms, which constitute

To account for the rise of the pronominal function of uomo, MeyerLilbke suggests French influence.2 But the general occurrence in Romance of homo with the value of an indefinite pronoun points rather conclusively to an antecedent development in Latin. Examination of the proper Latin authors reveals that homo was actually employed with pronominal force,3 and there is consequently little necessity of attributing the rise of this use of uomo to foreign influence. Uomo (< homo), like French on, was in origin a masculine substantive in the nominative case. We shall see to what extent these characteristic

PRONOUNS, which are always object expressions: syak 'I', FULL WORDS, which by themselves are open, but can be objectified by certain proclitic particles: simrek, PARTICLES, which are open and are never objectified: ken 'to, and', ti. 3. Object expressions are divided into two categories, STRAIGHT and OBLIQUE. The straight category is subdivided into NOMINATIVE and ATTRIBUTIVE, but this subdivision appears only in the enclitic pronouns. Object expressions are firstly, pronouns, and, secondly, full words and phrases OBJECTIFIED by certain proclitic particles. 4. ENCLITIC PRONOUNS appear in straight form only; they alone distinguish between nominative and attributive forms

ingenium, in Capt. 371 tute tibi tuopte ingenio prodes plurimum; thence to other abstract nouns, as in Trin. 313 istaec ego mi semper habui aetati3 integumentum meae and Bacch. 1083 nimi' nolo desidiae ei dare ludum; and finally to an abstract noun of action, as Bacch. 439 magistro desinebat esse dicto oboediens. There is reason to believe that partitive apposition was much commoner at an early stage of the language than later. It abounds in Hittite.32 As examples with the dative may be cited Al. 4.45 (Friedrich, Vert. 2.82) nu-kdn A-NA DUTU-SI SU-i an-da a-a?-?u lu-z~-lu a-ui 'and in my Majesty's hand (i.e. protection) behold goodly prosperity',33 Instr. 1.20 (Sturtevant, Chr. 148) nam-ma-kdn pi-di KA-as li-e ti-ya-zi 'let him not approach the place, the doors', i.e. 'the doors of the place'. Other cases are commoner than the dative. agent instead, as in Trin. 632 ut sic odio esses mihi. This exception seems to be the rule with one special type of noun of action, the verbal noun (really a dative supine) in -tui or -sui, as in Curc. 501 bono usui estis nulli. 29 Of course partitive apposition in Latin is by no means confined to the dative. See Hofmann's extremely interesting discussion and collection of examples, Syntaktische Gliederungsverschiebungen im Lateinischen infolge Erstarrung urspriinglich appositioneller Verhiiltnisse (IF 42.81-5). Hofmann cites numerous examples not only of the dative (84-5) but also of the nominative (85-6) and of the accusative (81-4), a few of the ablative (85) and of prepositional phrases (86), but none of the genitive; however, he suggests (85 fn. 2) that the construction lucis tuendi (discussed above, fn. 27) may exemplify this type. I formerly doubted this (see TAPA 74.282); but I now realize, as I have just said, that the fundamental relationship of the verbal noun to the substantive which it accompanies may be viewed as in origin one of quasi-apposition. 30 Some might term mihi a dative of agent, but actually the so-called dative of agent is essentially a dative of reference. 31 On the use of this particular noun in partitive apposition in the dative, cf. its similar use in the accusative, illustrated below, fn. 34. 32 For discussion and examples, see Friedrich, ZA 1.175 and 2.47, Vert. 1.43-5 and 2.24, El. 1.69; Sommer, Bil. 106. 33 It is hardly possible to render this literally. The force of the dative A-NA DUTU-SI must be 'with reference to my Majesty' and that of SU-i 'with reference to (my Majesty's) hand', but the adverb or postposition anda adds a locative notion 'in, within' at least to SU-i and perhaps to A-NA DUTU-SI also. (It must not be forgotten that the Hittite dative case serves as a locative as well.) This content downloaded from 207.46.13.131 on Sun, 16 Oct 2016 05:14:16 UTC All use subject to http://about.jstor.org/terms 250 LANGUAGE, VOLUME 29, NUMBER 3 Many constructions owe their genesis to early instances of partitive apposition. For instance, the accusative so used34 may have given rise to the so-called accusative of specification and accusative used as object of the middle voice3" met both in Greek authors and in Latin poets (like Vergil) who imitate Greek. Of course such an accusative must have started as the object of an active verb, but by a misunderstanding of the original construction remained as an accusative with a passive verb or an adjective, the transition possibly being achieved through the use of a middle verb, since this might be equivalent to an active verb with an accusative reflexive. The scheme of part and whole as we know it usually involves concrete nouns, generally denoting, as we have already mentioned, a person and a part of his body. But we have seen in our Latin examples the possible introduction, as the member representing the part, of an abstract noun, one signifying a quality or even, as in Bacch. 439, an action. If a verbal noun of the type of dicto can be placed in apposition with the substantive denoting the person uttering the dictum, why not, at an early stage, any verbal noun whatsoever with the substantive denoting the person or thing in some way connected with the action denoted by the verbal noun, whether as performing the action involved ('subject') or as receiving it ('object')? This possibility throws fresh light on all the Vedic infinitives and all the Latin gerunds. It is now easy to see why the construction that we have been studying is not confined to the double dative type. Just as Latin possesses a genitive and an ablative use of the gerund-gerundive construction, so Vedic possesses a combination of the genitive infinitive beside a genitive substantive and of the ablative infinitive beside an ablative substantive, parallel to the double dative.36 It is possible that the double accusative also enters into the picture. Vedic infinitives differ from Hittite infinitives in two important respects. (1) They include accusative forms37 as well as datives;38 these have two main uses, as the object of an auxiliary verb, like the Latin infinitive after it lost its dative force, and in expressions of purpose with a verb of motion, like the Latin ac34 As in Homer, Il. 7.14-6 'IUlvoov AXe bovpL t' ov, and Plautus, Rud. 1345-6 te Venus eradicet caput atque aetatem tuam. With the latter instance of apposition of part and whole, contrast their coordination in the same play 486 Neptuno credat sese atque aetatem suam; and for the appositional use of the accusative aetatem, cf. the similar use of the dative aetati in Trin. 313, already cited above (see fn. 31). 36 The constructions to which these names are given seem to me to be essentially the same, except that with a verb or a participle (e.g. Vergil, Aen. 1.561 voltum demissa) either name can be applied, and with an adjective (e.g. ib. 320 nuda genu) only that of an accusative of specification. It is significant that both usually indicate a part of the body, or, by an extension (as in Vergil, Ecl. 1.53-4 saepes apibus florem depasta), some portion of a whole. 36 For examples, see Gdr. 2.3.919. Brugmann adds to his Vedic instances parallel examples from Avestan, and then goes on to cite examples from Greek (920-1) and from Balto-Slavic (921-2). 7 Rather oddly, these are the only ones that survived into the classical period. 38 In Hittite, there is a supine in -wan which is regularly classed as a suffixless dative (or locative), but which I have sometimes been tempted to regard as an accusative; it is restricted to a very limited use in combination with not more than two or three auxiliary verbs, in the sense of 'begin'. This content downloaded from 207.46.13.131 on Sun, 16 Oct 2016 05:14:16 UTC All use subject to http://about.jstor.org/terms SOME HITTITE-SANSKRIT PARALLELS 251 cusative supine in -tum, with which the Sanskrit accusative infinitive is cognate. (2) They take an accusative object with far more freedom than the Hittite infinitive. Is there perhaps a connection between these two phenomena? Is it possible that the use of the accusative as the object of the infinitive had its genesis in an accusative substantive used in partitive apposition with an accusative infinitive, whence it spread to infinitives in other cases? Thus paian vicrtari vettha sarvdn, translated by Whitney (352) 'thou knowest how to loosen all bonds', would have originally meant 'thou knowest all bonds, (their) loosening', etc. There are still further implications. Since the verbal noun, being originally voiceless, may signify either an action performed by the accompanying substantive or one received by it, in other words since the substantive may have the relation of either subject or object to the verbal noun, we have the possibility that the double accusative construction may have a bearing on the development in Greek and Latin of the infinitive with subject-accusative.39 This is in line with the view of Ernout and Thomas, who believe that both the substantive that later became the subject of the infinitive, and the infinitive itself, were originally objects of the main verb. Thus they analyze sentio eum venire as a combination of sentio eum and sentio venire (271-2; cf. 274). However, on the whole this hypothesis seems to me dubious; would we not expect sentio veniendum (gerund)40 rather than sentio venire? It must not be forgotten that the Latin (and Greek) infinitives were originally datives, not accusatives; they show their dative force clearly in the construction iubeo (sino, cogo) eum venire, and it does not seem likely that they would have lost it so completely at so early a period in the construction scio (credo, dico) eum venire. Furthermore, though eum can be regarded as a possible object of scio,41 it seems hard to envisage it as an object of credo, and still more so as an object of dico; dico eum simply does not make

There are in Aleut' three endings to indicate the number of nouns: x' for the singular, x for the dual, and n (in the Atka dialect, s) for the plural. A'dax' father, a'dax two fathers, a'dan (Atka a'das) fathers. The singular has two cases, the absolute and the relative (or the nominative and the genitive). The absolute may serve in the sentence as subject or object (direct). All other relations between nouns and other words (with the exception of the instrumental and the comitative forms, for which see further) are expressed by the relative case and by postpositional pronouns added to the end of the noun. The dual and plural have only one case, the absolute, as the duality or plurality of the compound is expressed not in the noun but in the postpositional pronouns. The suffix of the relative case is m instead of the x' of the absolute case.

One of the most common items in the Albanian vocabulary, plak 'old man', pi. pleq, f. plake 'old woman', has never received a conclusive etymological explanation. Tagliavini (L'albanese di Dalmazia 222-3), following Jokl, properly rejects Meyer's connection with Gk. palaios. The latter must be related to pdlai, and this in turn, together with Skt. cirdh and Welsh pell, to IE *ktoel-; Pokorny accepts this doctrine (IEW 640). Like Jokl, Tagliavini and Pokorny (IEW 804-5) relate plak to Gk. pelios, polios, etc., under a base *pel-. This seems like a good fit. But, as we shall see, it is not necessary to follow Jokl in considering the final -ak a suffix from Slavic. Not only does the word bear all the appearances of belonging to the old layer of the vocabulary; there is no evidence that so tenuous an element as plcould have been segmented as a productive base to receive a borrowed suffix as late as the first millenium A.D. Furthermore, the conservative Greek enclaves show an initial cluster pl-, with no trace of an intervening vowel. Although we find Gk. pelitnos beside Skt. palitd-, the existence of Gk. pelios, polios, p6leia makes the iota appear not necessarily old. Hence palitdmay equally well be interpreted as a set base. Armenian alik' 'waves, white beard,' points to an apparent *p1iio-; in IE terms this is more effectively interpreted as *plXio-. The presence of the laryngeal (X) then readily explains how the *I syllabified. The Avestan and modern Persian cognates may well point to a similar syllabic background. Thus, without even assuming (as we are still free to do) a laryngeal suffix, we arrive at a seemingly attested sequence *p(e)lX-. Now, Martinet has recently offered, in BSL 51.42-56 (1955), the ingenious and attractive suggestion that many apparent instances of a 'suffix' in -kin IE are really nothing more than the result of the conversion to a velar stop of a laryngeal before s. Albanian plak looks like another such case. Thus all features are accounted for if we posit an old nominative (later generalized as a stem) *pldks, and an old accusative *pleAm. These would be from still earlier *plMAs, *pleAm. In similar fashion the entire entry (IEW 831) *pla-k-, *pla-k-, etc. can be brought under *pala(IEW 805-6). Gk. pkdx, plakos 'flat surface' is then from *pldks, *pl6Am, from earlier *pleAs, *plAAm, *plA6s. In view of Hittite palfyis 'wide', we may be more precise about the laryngeal: *plxs, *plExm, *plxis. If this is so, it seems likely that the word for 'gray' (which we provisionally assume not to have been a homonym) was *plehs, *pldhm, *plhEs. We may now proceed to an interesting point of chronology. Because of the sporadic occurrence in various dialects of forms such as Gk. n6ax, Lat. audax

Among the modern Romance languages, Roumanian is unique in possessing certain case endings for nouns and adjectives (including the articles). Each such word may have, in the singular and in the plural, three forms: nom.-acc., gen.-dat., and vocative. Very often the vocative is identical with the nominative; but there are four other endings: masc. sg. -e, -ule; fern. sg. -o; pl. -lor. Many writers have explained one or more of these vocative forms as borrowings from Slavic neighbors, in particular from Old Bulgarian; this explanation is almost invariably given for the feminine ending -o. I wish to present certain arguments against this theory. With such an object in view it will be instructive to consider the origin of all the vocative endings of Roumanian. To take our four endings in ascending order of difficulty, I will begin with the pl. -lor, as in Domnilor ?i doamnelor 'Gentlemen and ladies' (beginning of a public address). This form, which is of moderately rare occurrence,' is not even discussed by any authority that I have been able to discover. Apparently all scholars take it for granted-rightly, no doubt-that the form is really what it seems to be, identical with the articulated gen.-dat. plural in -lor, and thus of native Roumanian, i.e. Latin, origin (-lor < Lat. ill0rum). I do not know of another instance where an oblique case is used for a vocative; but one might compare examples of oblique-case forms used as nominatives, such as Italian Lei, Loro, Eng. you, thee (Friends' plain talk). In the masculine singular there are two endings, -e and -ule, both of which may, in many instances, be used for the same word. Of -e Tiktin2 says, 'Von diesen Formen stimmt die auf -e zwar zum lat. Vokativ (domine), ist aber wohl eher slavischer Herkunft (rabe Vokativ von rabti)'; similar statements are made by Meyer-Liibke3 and Bourciez.4 However, I think the correspondence between Roum. Doamne5 and Lat. domine, and numerous similar pairs, is too striking to be disregarded. The ending -e (or -ule) occurs in precisely the same type of words as in Latin-those ending in -u or a consonant (where -u has been

which are characterized by m, n (written -.m when final), and s (alternating with s). Compare SSS' 168 ff., who rightly maintain that these three paradigms go back to a single kernel. It is also clear that this kernel is the IE stem *to*tdwith s-forms in the nom. sing. masc. *so, fem. *sd. On the other hand the interpretation of the three characteristic consonants as being added particles (SSS 28 f.) cannot be upheld when it is realized that some of these forms correspond in their entirety to IE forms of the same meaning. Thus tdm, the fem. oblique of the first series, is identical with the IE acc. fem. *ta-m; and to?,2 the nom. pl. fernm. of the third series, is the IE *tas of Skt. tdh, Goth. frss, Lith. tos. Instead of seeing particles in these elements -m and -s, which from the Tocharian point of view seem secondary additions, we must therefore interpret as IE inheritances as many of the entire forms as possible, and then the others can be understood as formed analogously to these. However, the more precise history of these paradigms can be understood only in the light of the history of the Tocharian nominal declension, for the demonstrative paradigms are the result of the influence of Tocharian nominal paradigms on the inherited pronominal forms of the nominative and accusative (Toch. oblique). Bearing in mind therefore that also the cases of Tocharian nouns are divided into primary inherited cases (nom., obl., gen.), and new cases made by the addition of prepositions or particles (SSS 36), we shall deal with the demonstrative pronouns analogously. First will be considered the history of the nominative and

Although I believe that the Finnish so-called long vowels are phonemically geminate clusters, I think that the term 'short-vowel stem' is a convenient designation for declensional or 'conjugational stems ending in a single vowel. In the following, when I speak of me-stems, se-stems, te-stems, etc., these are to be understood only as varieties of e-stems. Every declined word in Finnish has a vowel stem to which all or most of the case endings are suffixed in the singular. Many words also have a consonant stem; many do not: neither type can properly be said to predominate. Those words that do have a consonant stem form the partitive singular on that instead of on the vowel stem, occasionally also an alternative essive singular form. The other words build the entire singular paradigm on the vowel stem, except, sometimes, for the nominative. Among dissyllabic stems this exception applies only to e-stems and pronouns. The plural declension need not be considered in this paper. If there is no consonant stem in the singular there is none in the plural. Every vowel in the language occurs as stem vowel in the dissyllabic shortvowel stems. Except among the e-stems, there is no consonant stem; and (neglecting the pronouns) the nominative singular is the same as the stem: paita 'shirt', genitive paidan,' partitive paitaa; hdntd 'tail', hdnndn, hantad; viikko 'week', viikon, viikkoa; tytt6 'girl', tyt6n, tytt6d; lintu 'bird', linnun, lintua; myrsky 'storm', myrskyn, myrskyd; nyrkki 'fist', nyrkin, nyrkkid. Some two-syllable given names in -e and a very few other dissyllabic e-stems follow this pattern exactly: kolme 'three', kolmen, kolmea. So Aune, Kalle, Roope, etc.; also itse 'self', nukke (frequently irregular in the plural) 'doll', kaase (obsolete or extremely rare variant of kaaso) 'bridesmaid'. And there are about a hundred of the dissyllabic e-stems that follow the same pattern except for a change of final vowel to -i in the nominative: sormi 'finger', sormen, sormea. The rest use a consonant stem for the partitive, like the polysyllabic e-stems, or else they show variation between the two types of declension. The polysyllabic e-stems form a very large and important part of the Finnish vocabulary. Most of them are derived from shorter words by means of formative suffixes. They all have a consonant stem, which may or may not be identical with the nominative. Examples are joutsen 'swan', joutsenen, joutsenta;2 hapan 'sour', happamen, hapanta; vdsymys 'fatigue', vdsymyksen, vdsymystd

1. Analysis of the grammatically complete sentence in Ukrainian reveals, as elsewhere in Indo-European, the familiar 'logical' polarity of subject and predicate. The nucleus of the subject is preponderantly a nominal element, that of the predicate a verbal one; apart from their characteristic suffixes of derivation, the first is defined morphologically by a system of declension, the second by a system of conjugation, i.e. by two contrasting patterns of flexion. The fact of flexion itself serves as a criterion to discriminate between the inflected and the uninflected lexical constituents (lexemes) of the language. The nominal system is, if anything, rather more involved than the verbal and covers the expression of a series of notional categories, only some of which (e.g. person and number) it shares with the verb. Selected case-relations are represented by more or less characteristic morphemes, and the variety of these is multiplied by the demands of gender and number. The clarity of the Ukrainian system of declension is somewhat obscured by the intrusion of the category of gender into a slowly disintegrating complex of variant paradigms. The predominance of gender inflexion is so obvious that synchronic grammar prefers a generic classification of declensional types to the earlier diachronic classification by stems defined partly in hypothetical terms (o-declension, d-declension, etc.). Nominal flexion implies the expression of several sets of correlations of varying scope: NUMBER, pointing a contrast between singular and plural; GENDER, here a complex of four antitheses (animate/inanimate, personal/impersonal, feminine/non-feminine, masculine/neuter) expressed in three overlapping generic groups; SYNTACTIC POSITION, opposing attributive to predicative function; and CASE, which presents the dichotomy of nominative (casus rectus) and oblique (casus obliquus), the former regarded as fundamental and amorphemic (merkmallos'), the latter embodied in a paradigm of semantically associated endings. The category of number shows vestiges of a dual, which has complicated the syntax of the genitive singular without affecting the correlation of singular and plural. This antithesis, however, is not uniform, as it does not possess the same connotation in the substantive and its immediate congeners on the one hand and the personal pronoun on the other (see below), and even in the substantive the plural may be conceived as either collective or multiplex. Gender, notionally, is more than the customary grammatical trinity of masculine, feminine, and neuter; it represents the intersection of the categories of animateness, person, and sex with their characteristic polarities. As for the case paradigms, they offer only a few instances of the correlatives nominative and oblique (e.g. sto

1 The footnotes in this article may seem extensive; their length is determined by the state of the study of Indo-European morphology. At the end of the nineteenth century many Indo-Europeanists were attempting to reconstruct older stages of Indo-European, but most of their proposals met with little success. The classical instance of such neglect is Saussure's assumption of laryngeals in 1878, which was disregarded until 1927, when Kuryiowicz demonstrated reflexes of laryngeals in Hittite. To incorporate all previous proposals in an article would confuse the course of exposition and would disturb specialists through its repetition of familiar data. On the other hand, omission of all reference to them would be unjust and might occasion difficulties for nonspecialists who become interested in consulting older materials. My notes will therefore survey materials which now have significance merely for the history of Indo-European studies and for the data they contain. Some footnotes will of course fulfill the common role of discussing matters germane to the text but of secondary importance in establishing the matter in hand. Among the fundamental studies are the following. In Die Pluralbildungen der indogermanischen Neutra (Weimar, 1889), Johannes Schmidt studied the various methods of forming nominative/accusative plural neuters in the Indo-European dialects, found that none of them were distinct from methods of forming nominative singular feminines, and concluded that in Proto-Indo-European neuter plurals were actually singular collectives. His conclusions have often been repeated and reaffirmed, as in J. Kurylowicz, L'apophonie en indo-europeen 58 and 84 (Wroclaw, 1956). Kurylowicz says (84): 'On sait depuis le travail celebre de J. Schmidt (PdiN) que le pluriel du neutre provient d'une transformation d'un abstrait (> collectif) du genre feminin.' Schmidt showed that the two forms were alike in origin; his demonstration did not establish the presence of feminine gender in preIndo-European. For a discussion of the faulty methodology which leads to such a conclusion see ?2 of this article. In Compositum und Nebensatz (Bonn, 1897), Hermann Jacobi reviewed the oldest types of compounds in Indo-European, and against a background of examples from many languages assumed that nouns at an early stage of Indo-European had not been inflected. He assumed that congruence classes had their origin in natural gender, the nominative singular in the pronoun so; despite these untenable assumptions, his short book is still of interest. In Der nominale Genetiv singular im Indogermanischen in seinem Verhaltnis zum Nominativ (Zwolle, 1902), N. van Wijk reviewed the various methods of forming genitives. He concluded that the genitive and nominative singular of consonant stems were alike at an early stage of Indo-European. 2 K. Brugmann treated the origin of gender in a number of articles and books, e.g. Zur Frage der Entstehung des grammatischen Geschlechtes, PBB 15.523-31 (1891), and The nature and origin of the noun genders in the I.E. languages (New York, 1897), summing up his views in Gdr.2 2.2.1.82-109 and in KVG 354-62. It was one of his achievements, in the argumentation surrounding his writings on gender, to carry through the view that the same linguistic conditions obtained five thousand years ago as today. In insisting on this principle in his lecture at Princeton in 1897, Brugmann gracefully cited Whitney's statement