Dėl naujų publikacijų lietuvių kalbos teorinės fonetikos ir morfologijos klausimais

ACTA LINGUISTICA LITHUANICA XLVII (2002), 117–130 REMARKS Remarks ON NEW PUBLICATIONS ON THEORETICAL PHONETICS AND MORPHOLOGY OF THE LITHUANIAN LANGUAGE The specialization of the two main periodicals of the Institute of the Lithuanian Language has recently begun to emerge: the journal Archivum Lithuanicum, launched a few years ago, is devoted to the history of written Lithuanian and the philological and textological analysis of older texts, while ALL (Questions in Lithuanian Linguistics), which will soon celebrate its half-century anniversary, is becoming more focused on theoretical and descriptive research into the Lithuanian language. Among the publications in the latest (46th) volume of ALL, which describe mainly contemporary Lithuanian material from various perspectives, two articles devoted to theoretical questions stand out: Kazimieras Garšva’s “Phonological Interpretations of Lithuanian Speech Sounds” (pp. 1–17) and Artūras Judžentis’s “A New Look at the Category of Gender in Lithuanian Nouns” (pp. 39–47). It is always useful to return to the discussion of unresolved questions. However good the Academic Grammar of Lithuanian may be, it too can be improved and developed, and the description of the language modified in light of new material and developments in linguistic theory. The seminars discussing the New Concept of Lithuanian Grammar are also devoted to this purpose; they are held at the Institute of the Lithuanian Language. Thus, the authors of the aforementioned articles are very much to be commended for their wish to take a fresh look at the interpretation of certain important aspects of the language system. Garšva boldly challenges certain claims that have become almost established (formulated mainly by Girdenis’s phonological school), on which the Academic Grammar of Lithuanian1 has also relied in recent years (since 1985), and proposes a new outline of a substantial fragment of the Lithuanian phonological system. At times the author agrees with tradition, at times he does not. In both cases he presents arguments, although many of them have long been familiar, and the examples he cites have mostly already been analyzed in earlier works by phonologists. Unfortunately, the author does not set out his theoretical principles in much detail, stating only that he draws on the Prague, Moscow, and Lithuanian schools of phonology. The unconventional interpretations presented in the article without particular comment are therefore surprising. It is unclear whether these are important parts of the author’s conception, worth debating, or merely infelicitous formulations. For example, it is quite strange to speak of the length of FEATURES as the basis for distinguishing segmental and suprasegmental units (p. 1). Further on, the discussion does refer to the length of phonetic UNITS (“Phonetic units longer than a phoneme are called suprasegmental..., those shorter than a phoneme—subsegmental”), but this distinction between prosodic and segmental elements is not customary either. One might infer that a phoneme is not the smallest element of the expression plane of language, since there are elements shorter than it (according to Garšva 1 Kazimieras Garšva co-authored the Phonology chapter of the Academic Grammar with Alekso Girdenis. 118 | PASTABOS these are moras2. In fact, the essential distinction lies on an entirely different plane: not in length, but in whether the element is linear or simultaneous (see Girdenis 1995: 38). Unfortunately, there are quite a few infelicitous formulations in the article overall; compare the recommendation to pronounce the sounds “[e], [e] as [a], [a]” after palatalized consonants (p. 3)3; it is unclear what purpose is served by the tautological statements that “the vowels [i], [i], [u], [u] are long and half-long” (p. 4)—this is evident from the transcription itself. Some “odd” features may be explained by the inconsistent use of square and slanting brackets, which blurs the distinction between phonological and phonetic phenomena. For example, the article refers to “variants of the vowels [a] and [e]” (p. 2)—it would presumably be more accurate to use the symbols for the phonemes /a/ and /e/, since (allo)phones do not relate to one another as an invariant and its variants4 (cf. also: “the short vowels [e], [e], [o], [a] could be regarded as positional variants of the vowels [i], [e], [u], [a]...” (p. 5), “one peripheral phoneme (short [o])” (p. 14), etc.). Let us set aside such questionable formulations and consider the more important claims in Garšva’s article. One of the most complex questions for phonologists has always been whether diphthongs and other vowel sequences should be interpreted as independent phonemes or as combinations of phonemes. Different phonological schools apply different criteria for establishing bifonemic or monophonemic status. The interpretation of “suspicious sound complexes” in Lithuanian (Girdenis 1995: 74) depends not on successful searches for new material, but on the development of an adequate theory. Garšva himself acknowledges that the two most commonly applied criteria for deciding the mono-/biphonemic question—the syllable boundary and the commutation test—are not absolute5. In his view, division into two syllables can serve only as an additional argument for the biphonemic status of diphthongs, since “this also occurs with long vowels (lý-ti: lí-jo, žú-ti: žù-vo)” (p. 7)6. He assesses the principle of commutation similarly: “Commutation (replacement by other phonemes) of the parts of diphthongs is, in my opinion, a more formal, accidental, and therefore less essential element in proving biphonemic status” (p. 8). Yet this does not prevent the author, in particular cases, from using now one, now another of the above criteria as decisive, or from dispensing with both. Thus, without relying in principle on new arguments or revising Girdenis’s theoretical principles (Girdenis 1995: 71–91), Garšva reaches the opposite conclusion regarding the biphonemic status of the Lithuanian diphthongs ie and uo7. This shows that theoretical premises allow researchers to choose the interpretation they simply prefer. 2 I wonder whether the author is inclined to regard a unit such as the syllable, which is longer than a phoneme, as suprasegmental as well? 3 The wording in the Academic Grammar is not much better, admittedly: “in the position [C⁻] /aː a/ are most often pronounced as /eː e/” (Ambrazas, ed., 1994: 24). 4 Although this is not customary in the work of Lithuanian phonologists, it would seem useful to distinguish terms such as BALSIS (masculine)—a sound, (allo)phone (written in square brackets)—and BALSĖ (feminine)—a phoneme (written between slashes). This would help formulate certain claims more precisely and sometimes avoid ambiguity. 5 Phonetic premises for biphonemic status (duration of pronunciation, not just articulation—see Girdenis 1995: 74) are generally ineffective: every phonetic (physical) feature is continuous, so segmentation could be based ONLY on an ARBITRARY AGREEMENT about the threshold value. 6 It is all the more strange, then, that the author uses just such alternation between tautosyllabic and heterosyllabic diphthongs as the basis for the biphonemic status of the diphthong uo. The relevance of the alternations duo-ti: da-vė, važiuo-ti: važiuo-vo to deciding this question is highly questionable in general—it is hardly purely phonological in nature (after all, as Garšva himself notes, in many cases the diphthong uo (and ie) is retained before a vowel: nuoalpis, prieangis; in secondary verbs the diphthong changes only before the past-tense stem-final vowel, whereas before the present-tense stem-final vowel a j is inserted to avoid hiatus). Moreover, the very possibility of treating tautosyllabic uo and heterosyllabic a-v as one and the same segment requires separate proof. 7 To be sure, he never does say which two-phoneme combinations these segments are supposed to represent (the variants mentioned include i+e, i+a, u+o, u+a (p. 8), and even a+v and u+v (p. 9)). PASTABOS 119 Indeed, both of the criteria mentioned were justifiably criticized as non-functional. It seems worthwhile to quote two longer passages from V. Kasevičius’s work General and Eastern Linguistics: Problems of Phonology (1983). Dėl komutacijos testo (plg. Girdenis 1995: 80): „... Asociatyvinė segmentacija [komutacijos testas – A. A.] fonologijoje – sąmoningai ar nesąmoningai – reprodukuoja [...] analogišką morfeminės analizės procedūrą, tačiau morfeminė analizė suponuoja sistemingą dviejų planų – išraiškos ir turinio – įtraukimą į palyginimą, ko visiškai neturi tik išoriškai analogiška fonologinė analizė. Iš tiesų, jeigu morfologijoje kalbininkas tenkintųsi tik morfemų eksponentų palyginimu, būtų korektiškos tokios proporcijos, kaip koza: kozël = oca: ocël, dėl to formą ocël tektų laikyti dvimorfiška, tai yra oc-ël. Tam trukdo oc reikšmių neatitikimas žodžiuose oca ir ocël, tiksliau – žodžio ocël segmento oc reikšmės stoka (asilas nėra vapsvos patinas, nors ožys yra ožkos patinas). Bet fonologinės segmentacijos, kuri pagrįsta asociatyvine analize, atveju kalbininkas tiria tik išraišką, neatsižvelgdamas į turinį. Todėl gali atsirasti nemotyvuotų sprendimų, kaip rodo ką tik pateikti morfeminio skaidymo pavyzdžiai neatsižvelgiant į turinį“ (Kasevič 1983: 22)¹. Regarding the syllable boundary (the assignment of individual parts of a segment to more than one syllable (Girdenis 1995: 74)): “Trubetzkoy classified Rule I as phonetic, considering the syllable itself a phonetic unit. However, it is precisely the boundaries of phonetic syllables that may sometimes occur ‘inside’ segments that are unquestionably phonologically unitary. Thus, in Burmese, when an onset consonant is strongly attached to the preceding syllable, conditioned by that syllable’s tone, this consonant becomes two-peaked. Phonologically, it remains within the following syllable, but its two peaks indicate its phonetic distribution across two syllables⁸.” If we assume that the concept of the syllable in Rule I can be interpreted from a phonological perspective, then in this case the syllable should be understood as a certain pattern of phoneme organization, where the boundaries between syllables are, naturally, also boundaries between phonemes. But then, to use syllable boundaries in phonemic segmentation, one must have a comprehensive definition of the syllable in terms of its phonemic organization. Such information is hardly available at the very outset of phonological analysis” (Kasevič 1983: 31)⁹. * “...Associative segmentation in phonology—consciously or unconsciously—reproduces [...] a procedure analogous to morpheme analysis; however, morpheme analysis entails the systematic comparison of two planes—expression and content—which the merely externally analogous phonological analysis entirely lacks. Indeed, if in morphology the linguist were content to compare the exponents of morphemes, proportions such as коза: козёл = оса: осёл would prove valid, and consequently the form осёл would have to be represented as bimorphemic, i.e. ос-ёл. This is precluded at least by the lack of meanings for ос in оса and осёл, or, more precisely, the absence of a meaning in the latter (a donkey is not a male wasp, although a he-goat is a male goat). But in phonological segmentation based on associative analysis, the linguist deals only with the plane of expression, without recourse to the plane of content. Hence the possibility of unmotivated decisions, demonstrated above using morpheme division not supported by reference to the plane of content.” 8 Among matters closer to Baltists, one might perhaps mention the Latvian voiceless plosives, which are palatalized between two short vowels: does the division of a long consonant by a syllable boundary in the word lapa [lap-pa] require us to distinguish two phonemes p: /lappa/? – A. A. ⁸⁸ “Trubetzkoy classified Rule I as phonetic, considering the syllable itself a phonetic unit. However, the boundaries of phonetic syllables may sometimes be found ‘inside’ segments that are unquestionably phonologically unitary. Thus, in Burmese, when an onset consonant is strongly attached to the preceding syllable, conditioned by the tone of that syllable, this consonant becomes two-peaked [82]. Phonologically, it remains within the following syllable, but its two peaks indicate its phonetic distribution across two syllables.” 120 | PASTABOS It is odd that Lithuanian phonologists make almost no use of the morpheme-boundary criterion when addressing the problem of mono- versus biphonemic status (Girdenis (1995: 90) does not mention it either; see Čekmonas (1983: 199); but cf. Vaitkevičiūtė (1961: 39); Pupkis (1966: 117) openly tends to disregard it). The morphological-boundary criterion follows directly from the traditional definition of the phoneme as the smallest FUNCTIONAL segment of the speech stream (cf. Ambrazas, ed., 1994: 15). Since phonemes constitute the expression of morphemes, it is clear that a morpheme boundary identified within a phonetic segment unequivocally attests to a phoneme boundary as well⁹. Thus, to demonstrate that a suspect segment is biphonemic, one must find an example in which a diphthong is divided by a morpheme boundary. This would require a separate study, but for now only a few preliminary observations can be made. If we adopt the segmentation of finite verb forms proposed in the Academic Grammar (Ambrazas, ed., 1994: 332–333), in which the thematic vowel is assigned to the stem, then in first- and second-person forms with long stem-final vowels the diphthongs au, eu (iau), ai, ei arise precisely where a morpheme boundary (between stem and ending) occurs. If we follow the more traditional segmentation of these forms, the segments mentioned belong to the ending. In that case, it is difficult to find an example of a morpheme boundary within a compound diphthong. Both variants of verb-form segmentation provide examples of the division of mixed diphthongs: cf. the stem-ending boundary neša-m, myli-m, kelia-m [em]; neša-nt, myli-nt, kelia-nt [ent] or the infix boundary ta-m-pa, gle-m-ba, čiu-m-pa, li-m-pa, ra-n-da, skre-n-da, ju-n-ta, kri-n-ta10. Diphthongs with liquids r, l can probably be analyzed by analogy. However, it seems impossible to find an example in which tautosyllabic diphthongs are divided by a morpheme boundary. The morpheme-boundary criterion is also applicable to determining the phonological status of affricates, on which Garšva does not make his position entirely clear. It is true that there is no consensus regarding the formation of affricates at morpheme boundaries (at-sikėlė, pať-s); experimental studies are still incomplete, and their findings are contradictory (see Pakerys 1995: 156 (with references), 163). Besides the phonological interpretation of complex segments, it is worth noting several other points discussed in Garšva’s article. The author’s understanding of the mora seems unusual. The previously mentioned treatment of the mora as a subsegmental unit—shorter than a phoneme—is already questionable (p. 1). Moreover, it also contradicts the principle of mora theory formulated somewhat later by the author himself: “According to mora theory, long vowels are a combination of two identical short phonemes (moras), equivalent to one syllable” (p. 6)—yet this implies that a mora is not shorter than a phoneme. Most importantly, however, the equation of a mora with a short vowel is itself unjustified. The fact that a long vowel (more precisely, a syllable) in mora-timed languages functionally corresponds to two short vowels does not at all mean that it consists of two short pho- If, on the other hand, we assume that the concept of the syllable in Rule I can be interpreted from a phonological perspective, then in this situation the syllable should be understood as a certain model of phoneme organization, where the boundaries between syllables are, naturally, boundaries between phonemes. But then, to use syllable division in phonemic segmentation, one must have a comprehensive definition of the syllable in terms of its phonemic organization. Such information is hardly available at the very outset of phonological analysis.” 9 In other words, “phonological segmentation proves to be derived from morphological segmentation” (Kasevič 1977: 34). This is one of the fundamental principles of L. Ščerba’s phonological school. 10 It is worth noting that the stem’s morphemic structure also depends on the variant used to segment verb forms: the stem suffixes -j-, -st-, -n-, -d- and the infix traditionally distinguished lose their significance under the segmentation adopted by the Academic Grammar, and thus also lose their morpheme status (stem variation is described in morphonology as alternation between allomorphs) (Andronov 2000: 44). P A S T A B O S | 121 … Moras are not directly related to phonemes in general. A mora should be regarded as an abstract, conventional measure of syllable length. If such a measure proves necessary, or simply useful and convenient, for describing a language system, the language is considered moraic. The arguments raised by Girdenis (1995: 259–260) and repeated by Garšva (pp. 6–7) against a moraic interpretation of Lithuanian are insufficient. If the mora is accepted as a conventional unit (cf. Girdenis 1995: 259), there is no longer any need to seek short counterparts for the long phonemes /oː/, /eː/ (Girdenis 1995: 260). The phonetic realization of pitch accents is not decisive in phonology. As for the relationship between pitch accents and stress, it would be worthwhile to examine in greater detail, for example, Jurij Stepanov’s theory (Stepanov 1972, 1997 a, 1997 b), which Garšva mentioned only briefly. In discussing the correlation of palatalization and velarization in Lithuanian, Garšva follows Girdenis and offers no independent arguments. Nor is the conclusion new: he proposes retaining the traditional interpretation (which assigns this correlation to consonants), on the grounds that other solutions would supposedly complicate language practice. It could be quite useful to demonstrate and assess those potential complications in detail (cf. Levin 1994). One cannot agree with the author’s explanation that “Latvian consonants are not divided into hard and soft because there are few of the latter” (p. 12). What matters is not the number of such consonants (how many would there have to be?), but the fact that they do not form correlational pairs with other consonants (see Trubeckoj 1960: 93–96, 151–153 = Trubeckoj 2000: 89–92, 145–147; Girdenis 1995: 140–151; on Latvian consonants, see Andronov 1998: 140–141). In the absence of a hard–soft correlation, it is generally inappropriate, from a phonological standpoint, to call palatal Latvian consonants “soft,” since that would mean imposing features of another language’s phonological system on Latvian. The question of the contrast between the phonemes /a/ and /e/ after palatalized consonants (pp. 4–5)11 is certainly worthy of attention. The three-volume Academic Grammar asserts that the phonemic contrast is generally preserved in this position (Ulvydas, ed., 1965: 55–56) (the exception is the diphthong ei, generalized after all palatalized consonants except the affricates č’, dž’; Ulvydas, ed., 1965: 58). This view is based on Vaitkevičiūtė’s general experimental studies of the phonetic characteristics of vowels (1961: 36–38, 43–44). The current Academic Grammar, however, adopts the view that only the phoneme /e/ is pronounced after palatalized consonants (Ambrazas, ed., 1994: 24). This view is also presented in Pakerys’s phonetics textbook for higher education (Pakerys 1995: 34, 35, 42, 154) and in Girdenis’s monograph on theoretical phonology (Girdenis 1995: 54, 62, 172)12. Unfortunately, these scholars cite no more recent experimental studies to support this view. Girdenis mentions only Kazlauskas’s statements (Kazlauskas 1967: 238; 1968: 325 = Kazlauskas 2000: 239, 283) and his own critical comment in his review of the 1965 Academic Grammar (Girdenis, Žulys 1973: 207 = Girdenis 2000: 376). Yet all this seems more like individual impressions of a language user than objectively established facts. To be sure, in support of his interpretation, Girdenis also cites rhymes from classical poetry: “Vincas Mykolaitis-Putinas rhymes not only girią: vakarė..., rūkuosė: dvasią..., but also teisės: baisios..., stalėlio: karaliai..., dėngia: padangė..., galiai: kėlia...” (Girdenis 1995: 172). It is unlikely that this circumstance explains anything: if a poet considers even the basic 11 Garšva is entirely right to point out that “the pronunciation of the standard language and the description of morphonological phenomena both depend on the resolution of this question” (p. 5). Akselis Holvoetas had already drawn attention to this in his review of the Academic Grammar in English (Holvoet 2000: 211). 12 Keista, kodėl nei Akademinė gramatika (Ambrazas, red., 1994: 22), nei Pakerys (1995: 150) neįtraukia į bendrą lietuvių kalbos dvibalsių sąrašą dvibalsio eu, tariamo, pagal jų aprašymą, po minkštųjų priebalsių (Girdenis (1995), rodos, nepateikia bendro dvibalsių sąrašo). 122 | PASTABOS difference between the allophones of the phonemes /a(:)/ and /e(:)/, how can his poems be used to infer an opposition between weakly differentiated allophones after palatalized consonants? Apparently, new experimental studies would be needed to clarify the situation (cf. Pakerys’s comment (Mikalauskaitė 1975: 6)). Garšva’s discussion of the contrast between the phonemes /a/ and /e/ after palatalized consonants also draws on the results of his own study. However, the methodology of the author’s survey is not described in detail in the cited article (Garšva 2001), and it appears that informants had to assess their own pronunciation, so their answers may have been quite subjective. In designing further experiments, it would be worthwhile to distinguish positions after palatalized consonants from those after the phoneme /j/, which does not belong to the hard–soft correlation (e.g., siaubia and siaube – jai and jei): preservation of the contrast after /j/ seems more reliable (cf. Mikalauskaitė 1975: 58; Zinkevičius 1966: 51). To sum up, Garšva’s article makes an unusual impression: it is very interesting to read because the author’s doubts seem entirely justified, but his proposed solutions are no more convincing than the claims he criticizes. The article is valuable precisely because it highlights not only questions concerning linguistic data that merit discussion, but also problems with the theory itself. One can only hope that this might encourage more active discussion among phonologists (such discussion has objectively been limited to date also because many phonologists’ works on standard Lithuanian and its dialects are published only in Lithuanian). Quite different thoughts arise in connection with Artūras Judžentis’s “new look at the category of gender.” First of all, it is not clearly explained here why the author is dissatisfied with existing descriptions of noun gender. At first glance, therefore, the article resembles an elegant game of glass beads (what might emerge if we describe everything quite differently?). The author’s approach seems entirely possible, but the central question is whether such a new description of the category of gender is appropriate—whether it is simpler and more universal, and whether it helps reveal phenomena or regularities not identified before. The main shortcoming of many descriptions of noun gender is that the CRITERION for determining gender (based on the form of agreeing words) is confused with, or presented alongside, the FEATURES of words belonging to each gender (both formal and semantic; cf. Bulygina, Stepanov 1988)13. Judžentis’s article opens in a similar way: “The gender category of Lithuanian nouns is expressed by their inflectional forms and their relationship to agreeing words” (p. 39; cf. a similar statement in the article’s conclusions, p. 45). Formulated by the author on the basis of the authoritative sources cited, this statement is ambiguous: does it assert (1) that in order to determine a noun’s gender we must take account of both its inflectional forms and agreement (a given noun belongs to gender G only if it has a P-type ending paradigm AND requires an F-type form from agreeing words), or (2) that either one of these factors is sufficient (a given noun belongs to gender G if it has a P-type ending paradigm OR requires an F-type form from agreeing words 13 An analogy may be drawn: the classification of nominals by accentual paradigms is carried out by a very clear procedure (using what is called at school the “M-letter rule”)—by analyzing the location of stress in the plural dative and accusative. However, various patterns in the distribution of nominals among accentual classes have been identified (it has been observed that primary adjectives do not belong to accentual classes 1 and 2, and the accentuation of derivatives can be related to the accentuation of base words or to the accentual properties of the derivational formant, and so on). Thus, for example, the description of the accentual classes of suffixed nouns could be presented as in Laigonaitė (1978: 41–46): each accentual-class section identifies the suffixes that form nouns belonging to that class and indicates exceptions. But this does not mean that one can claim that accentual classes are determined by suffixes. PASTABOS 123 requires a form of type F)? It seems that rules of the first type are not used by anyone in determining the gender of Lithuanian nouns. Thus, experience leads us to choose the second interpretation of the statement under analysis. However, if any one criterion is sufficient, it remains unclear whether there are cases in which one criterion contradicts another and, if so, how the researcher should proceed in such a case. Cases in which, to use Judžentis’s words, the inflectional forms of a noun do not correspond to its relations with the agreeing words (vaidila, akiplėša and others) are well known and described in the literature. In all such cases, the gender of the noun is determined by the form of the agreeing words—there are no cases in which inflectional forms are given priority here¹⁴. What, then, is the point of mentioning inflectional forms as one of the criteria for determining gender? Having begun his article with a dubious claim, Judžentis proposes in the very next sentence that the morphological and syntactic aspects of the gender category be distinguished. In fact, the traditional characterization of gender attempts to combine two categories: grammatical gender and morphological gender, or, more precisely, declension type (a third category is sometimes included as well—lexical meaning, or, more precisely, biological sex). The concept of morphological gender is known in Lithuanian linguistics mainly thanks to Bulygina’s work (1970: 26–33), but, as Judžentis notes (pp. 40–41), it is “insufficiently brought into focus” and has at least not been fully “utilized” in the description of the language. Judžentis does not provide a definition of the concept of morphological gender; instead, he specifies which declension classes’ nouns belong to which gender (pp. 39–40). In grouping the paradigms into two classes, he clearly relies on the grammatical gender of nouns, since otherwise it is unclear why precisely two morphological genders are distinguished, or why 12 paradigms are reduced to two classes. Morphological gender is a property of a paradigm, indicating for nouns of which gender the set of endings characteristic of that paradigm is used; the logic behind the emergence of this concept is set out by Zaliznjak (1967: 146–147). In Lithuanian, the morphological gender of nouns is unambiguously indicated only by the singular dative ending (Bulygina 1970: 28), so the morphological genders could be called “masculine, or ui gender” and “feminine, or ai/ei gender.” This distinguishing feature alone is sufficient to indicate when describing the difference between masculine and feminine declension types (cf. the long list of declensions from the Academic Grammar cited by Judžentis (p. 40)—Ambrazas, ed., 1994: 63–64). In addressing such a practical task as classifying declensions, it would also be preferable to rely not on the singular genitive¹⁵ but on the same singular dative form. It is unclear how Judžentis determines the morphological gender of pluralia tantum (p. 40)—their declension paradigms do not differ¹⁶. Apparently, he simply transfers grammatical gender to this context. 14 Continuing the analogy begun in footnote 13, we may note that it would occur to no one to exclude begėdis from the second accent paradigm on the basis of the familiar rule that derivatives with the prefix be- are stressed according to the second accent paradigm (Laigonaitė 1978: 46). 15 The use of the singular genitive form here can be explained only by tradition—by the influence of medieval Latin grammars: in Lithuanian, this form indicates neither morphological gender nor the inflectional stem (Ambrazas, ed., 1994: 68), which is important for classifying declensions (the plural dative indicates it most clearly). 16 A subtle distinction in the plural paradigm of the i-declension can be discerned only between the two main subtypes of this declension: the consonantal subtype (paradigms 11 and 12) and the i-declension in the narrow sense (paradigms 9 and 10). In the former, the genitive plural of nouns has a hard stem-final consonant, whereas in the latter it is usually soft (although there are exceptions). Recall that the traditional assignment of the noun durys to the i-declension in the narrow sense (paradigm 9—cf. Judžentis 2002: 40) is unfounded: because the genitive durų has a hard stem-final consonant, this word belongs here among the exceptions, whereas in the consonantal subtype it seems entirely regular. 124 | PASTABOS Noun endings do not have a grammatical gender morpheme; only the endings of agreeing words, for example adjectives, do (cf. Bulygina 1970: 32). Individual noun endings merely indicate morphological gender more or less clearly; in other words, they are features more or less characteristic of the masculine or feminine declension type. This is an entirely understandable conclusion given this conception of the grammatical gender category, yet it must be emphasized specifically because the opposite is often claimed. As we have seen, the grammatical gender of nouns is determined on the basis of the form of agreeing words. In summary, the category of grammatical gender is a certain inherent property of every noun, one that requires the selection of one form or another of an agreeing word and is not directly related either to form or to semantics (which also proves that gender is an autonomous category, rather than a formal declension type or an unambiguous reflection of reality). Since noun gender is a property of a part of speech (and thus a property of the object of morphology) that is realized at the syntactic level (through agreement between words), it should be regarded as a morphological-syntactic category. Defining the grammatical gender of a noun by its declension type or meaning is simply pointless: declension type indicates not grammatical but morphological gender (this does not apply to indeclinable nouns), and meaning is likewise related not to grammatical gender but to biological sex (this applies only to designations of persons and, in part, animals). A linguist’s task, therefore, is to describe the relation of the noun’s grammatical gender category both to word form (declension type, or morphological gender) and to semantics. From the morphological perspective, all nouns are first divided into declinable and indeclinable: the former are further classified as nouns of masculine or feminine morphological gender, while the latter have no morphological gender. From the semantic perspective, nouns are likewise divided into two subgroups: nouns in the first subgroup are further classified as masculine or feminine according to biological sex; those in the second subgroup are not classified, since they denote objects that do not have a distinction of sex, or for which that distinction is perceived in the language as irrelevant (cf. vaikas, briedis, varlė and others). Combining the grammatical, morphological, and semantic classifications yields 18 potential noun groups (2×3×3). However, cases in which grammatical and semantic gender do not coincide probably do not occur in Lithuanian17, while directly opposing grammatical and morphological gender without semantic grounds is also questionable. Thus, only 10 possibilities are realized (see Table 1). These thoughts arise from an attempt to improve somewhat the foundations of the traditional description of the category of noun gender. For further exploration of the gender category, the extensive discussions presented in several books by Aldona Paulauskienė, as well as the development of her own conception of the gender category, are highly valuable (Paulauskienė 1983: 172–183; 1989: 170–178; 1994: 153–166). We can now return to Judžentis’s theory of gender. Of great importance here is the explanation of the twofold nature of the agreement of words dependent on a noun: in some cases, they “agree according to the form of the head word”18 (syntactic agreement), and in others, “according to the sex of the noun’s referents” (semantic agreement) (p. 42). This recalls Valeckienė’s claim that “the semantic criterion 17 Except in cases of figurative use (Jis tokia boba) and colloquial formations (mielas Mildukas). 18 What is meant is specifically morphological form, or morphological gender; cf. the English abstract: “Inanimates have only morphological gender, which will also be the only factor determining the gender form of agreeing words (agreement is then purely syntactic)” (p. 256). PASTABOS | 125 TABLE 1. RELATIONSHIP BETWEEN GRAMMATICAL, SEMANTIC, AND MORPHOLOGICAL GENDER grammatical gender semantic gender morphological gender No. examples masculine masculine masculine 1. father feminine 2. uncle neither 3. messieur neither masculine 4. stalas none 5. taksi feminine feminine masculine 6. móteriškas ‘moteriškė’ (³DŽ 497) feminine 7. motina none 8. ledi none feminine 9. kėdė none 10. fojė as it includes precisely all those cases that the formal criterion does not mark (stem exceptions)” (Valec-kienė 1998: 242). Unfortunately, such a neat relationship does not actually exist: there are nouns to which no “criteria” apply except the syntagmatic one—these are plural nouns of the first declension (mostly its consonantal subclass), which lack a gender feature and inflect according to identical paradigms (delicious juices feminine: dangerous ironworks masculine; large livers feminine: terrible muscles masculine). If biological sex is absent, the gender motivation of indeclinable nouns can be sought by taking into account the gender of the kind of thing named (taksi – mašina / automobilis), or by the similarity of the ending to the inflectional ending of declinable nouns (fojė – žemė), but there is no unambiguous answer here (p. 44). This means that motivation need not necessarily exist. That is natural, since each criterion corresponds to its own category. Thus, there is in fact no complementary distribution relationship between the semantic and formal criteria (according to Valečkienė), or any semantic and formal (syntactic) alignment (according to Judžentis). If such a relationship existed, the grammatical gender category would not be an independent category, but only a secondary combination of semantic and morphological gender. Nevertheless, the explanation of semantic agreement is worth considering: according to Judžentis, nouns denoting persons and some animals agree semantically (pp. 42–43, 45). Following this line of thought, one could argue that the aforementioned group of words simply lacks the grammatical gender category—their dependent words agree with them according to the referent’s sex. In that case, such “exceptions” as masculine-gender words belonging to the feminine declension type would be eliminated (e.g., 126 | PASTABOS uncle), and substantiva communia19—if all these words lack grammatical gender, then there can be no problems of gender mismatch either. The advantage of this explanation is particularly evident in the case of indeclinable personal names in other languages (also a kind of substantiva communia): it is more economical not to assign a gender feature at all, for example, to the surname Kiuri20, and to explain the choice of gender in dependent words through semantic agreement with the referent’s sex. Unfortunately, the author’s position on this issue is not explicated in the article. This understanding of a “new perspective on the category of gender” naturally has certain drawbacks as well. Neither the noun class nor the gender category would now be uniform; moreover, if this approach were adopted, many other things would also have to be changed. For example, the internal division of the noun class becomes very important in this case, since different groups of nouns have different sets of grammatical categories21; a somewhat different definition of the noun itself and of the gender category itself would probably also be needed. In general, even the understanding of the syntactic relation could presumably no longer remain traditional. I do not know whether the author intends to reorganize the entire grammar and adapt it to this new understanding of gender. For now, taken on its own, the “new perspective on the category of gender” seems more like an attempt at pure theoretical construction than a focused investigation of the terrain of grammar. Judžentis also discusses specific issues in describing the gender of nouns, but proposes no new solutions. The author agrees with those linguists who tend to regard noun pairs mokytojas: mokytoja not as different lexemes, but as forms of a single lexeme inflecting for gender22. Since in Lithuanian person-denoting terms forming gender pairs are a very regular phenomenon, there is reason to assign them to the domain of inflection (cf. Krylov 2002: 734–736). However, apart from the aforementioned less-than-desirable division of the noun class, an argument against this solution is that the composition of the group of nouns inflecting for gender cannot be strictly delimited (terms for persons by profession, etc., as well as names of some animals), and the means used to form pairs are not uniform, cf.: liūtas: liūtė, žvirblis: žvirblienė, jautis: karvė. Moreover, if noun inflection for gender is recognized, one of the most important distinguishing features of the grammatical category systems of nouns and adjectives disappears: gender in nouns is a classificatory category, whereas in adjectives it is an inflectional one. A similar question is whether substantiva communia are better treated as two homonymous lexemes, as homonymous forms of different genders of a single lexeme inflecting for gender, or—as in the “new perspective”—as a word lacking a gender feature. There is another aspect here that merits attention: descriptions of Lithuanian grammar do not allow this group of words to be identified unambiguously. For example, the description in the Academy Grammar does not make clear whether substantiva communia (e.g., nevala) should be distinguished from nouns of one gender (usually zoonyms) used figuratively (e.g., kiaulė, boba, mazgotė), cf.: 19 Regarding the term, I fully agree with Bonifacas Stundžia: it denotes a noun class, not gender (Stundžia 1980: 87). 20 The traditional interpretation here posits two homonymous nouns (Kiuri masculine and Kiuri feminine), while the less traditional interpretation posits that such nouns inflect for gender. 21 Among nouns denoting persons and some animals (for which the sex feature is relevant), a further group is distinguished: terms for persons by profession and the like (mokytojas: mokytoja), which are proposed to be treated as inflecting for gender. 22 The author notes an inconsistency in the Academy Grammar in this respect: the gender of nouns is treated as a classificatory category (Ambrazas, ed., 1994: 62), but the section on word formation mentions noun inflection for gender (Ambrazas, ed., 1994: 142). PASTABOS 127 “In Lithuanian there is a sizeable group of nouns that can be used to name persons of either female or male sex on the basis of a negative trait characteristic of those persons. Their gender is expressed syntactically—by the forms of agreeing words—and depends on the sex of the person whom they denote in each particular case. Such nouns are conventionally called common-gender nouns” (Ambrazas, ed., 1994: 64). However, apart from a difference in secondary meaning (cf. Ulvydas, ed., 1965: 170), the distinction is also manifested in different agreement patterns: with nouns in the substantiva communia group, the form of the modifier is chosen according to the person’s sex (it is impossible to say he is a real slob), whereas with a “regular” noun used figuratively, agreement according to its “inherent” gender is also possible (he is a complete slob / she is a complete slob; she is a real pig / he is a real pig). In the first case we have genuine substantiva communia; in the second, words of masculine or feminine gender used only figuratively. A particularly interesting example found by Judžentis in the VDU Lithuanian Language Corpus is The highest-paid model Naomi Campbell looked rather tired: “In this case, the modifier agrees syntactically with the noun, whereas the predicative agrees semantically, according to the referent’s sex” (p. 44). It is difficult to draw conclusions from a single example (especially since it is not entirely clear: would a similar sentence be possible without mentioning the name: the good model is tired?). But in general the situation strongly recalls an example that has attracted considerable attention among Russian linguists: новый врач пришла (Bulanin 1976: 48–54). It seems that in both languages the syntactic relation between subject and predicate is more closely tied to semantics than is the agreement relation between a head word and its modifier (perhaps this is a universal phenomenon?). Semantic agreement in such a case is understood as a deviation from the norm, in which the syntactic relation links not linguistic units but directly reflects reality. Lithuanian, where such utterances are only beginning to appear, is once again more archaic than Russian, where in similar cases predicate agreement by meaning is sanctioned in grammars. Indeed, if we were to interpret in a similar way the previously mentioned fluctuations in the gender of dependent words accompanying nouns used figuratively (where agreement may be either grammatical or semantic), it would appear that semantic agreement is even more widespread in Lithuanian than in Russian. A more thorough comparison of the facts in the two languages would be worthwhile. In conclusion, it must be acknowledged that careful studies of individual aspects of the gender category are needed: it turns out that not all relevant material has been collected and classified. It is therefore still impossible to reach a definitive conclusion on whether to choose a traditional or a new description of noun gender. In any case, it should be borne in mind that changing our understanding of one component of a grammatical system will probably also require changes to the descriptions of other components. Other categories—not just gender—should be described, with morphological, syntactic, and semantic approaches rigorously distinguished. Many fruitful ideas for further research into Lithuanian morphology can be drawn from Zaliznjak’s detailed description of the inflection of Russian nominal words (Zaliznjak 1967; 2002), which S. Krylov examines attentively from a contemporary perspective (Zaliznjak 2002). Many questions that Judžentis reflects on, and that others often fail to notice altogether, have a long history in Russian linguistics and have prompted interesting conclusions. I would not want the impression to arise that I am biasedly promoting the theories of Russian linguists here. It seems to me that their importance for Lithuanian linguistics is objectively determined by the fact that the systems of Russian and Lithuanian are far more similar than, for example, those of Lithuanian and the Germanic or Romance languages (whose material has largely been described using theories of European and American linguistics). Russian has been described better than Lithuanian in many respects; critically assessing interpretations of some of its phenomena would help reveal subtle differences between the two languages and thus lead to a better understanding of them. 128 | PASTABOS The development of new linguistic approaches should not overshadow the study of traditional branches. With this in mind, one would like to see more publications on questions concerning the conception of Lithuanian grammar. LITERATŪRA AMBRAZAS, V., red. 1994: Dabartinės lietuvių kalbos gramatika, Vilnius: Mokslo ir enciklopedijų leidykla. ANDRONOV, A. 1998: A Short Grammar of Latvian by Terje Mathiassen, Columbus (Ohio) 1997. Res Balticae 4, 139–158. ANDRONOV, A. 2000: Some remarks on the system of Lithuanian and Latvian conjugation. Linguistica Baltica 8, 35–47. BULANIN, L. L. 1976: Трудные вопросы морфологии. Москва: Просвещение. BULYGINA, T. V. 1970: Морфологическая структура слова в современном литовском литературном языке (в его письменной форме). Žirmunskij, V. M., Arutjunova, N. D., red., Морфологическая структура слова в индоевропейских языках, Москва: Наука, 7–70. BULYGINA, T. V., STEPANOV, JU. S. 1988: Теория грамматики и теоретическая грамматика (К выходу в свет „Грамматики литовского языка“ на русском языке). Вопросы языкознания № 3, 17–33. ČEKMONAS, V. 1983: [Rec.] A. Girdenis. Fonologija, Vilnius, 1981. Baltistica 19 (2), 197–204. 3DŽ – Dabartinės lietuvių kalbos žodynas, vyr. red. S. Keinys, Vilnius: Mokslo ir enciklopedijų leidykla, 1993. GARŠVA, K. 2001: Bendrinės kalbos tarties problemos. Poljakovas, O., red., Žmogus kalbos erdvėje, Kaunas: Vilniaus universiteto Kauno humanitarinis fakultetas, 143–148. GARŠVA, K. 2002: Lietuvių kalbos garsų fonologinės interpretacijos. Acta Linguistica Lithuanica 46, 1–17. GIRDENIS, A. 1995: Teoriniai fonologijos pagrindai, Vilnius: Vilniaus universitetas. GIRDENIS, A. 2000: Kalbotyros darbai 1, Vilnius: Mokslo ir enciklopedijų leidybos institutas. GIRDENIS, A., ŽULYS, V. 1973: [Rec.] Lietuvių kalbos gramatika 1, 2, Vilnius, 1965, 1971. Baltistica 9 (2), 203–214. (= Girdenis 2000: 370–386.) HOLVOET, A. 2000: [Rec.:] V. Ambrazas, ed., Lithuanian grammar. Lietuvių kalbos gramatika, Vilnius, 1997. Linguistica Baltica 8, 210–214. JUDŽENTIS, A. 2002: Naujas žvilgsnis į lietuvių kalbos daiktavardžių giminės kategoriją. Acta Linguistica Lithuanica 46, 39–47. KASEVIČ, V. B. 1977: Элементы общей лингвистики, Москва: Наука. KASEVIČ, V. B. 1983: Фонологические проблемы общего и восточного языкознания, Москва: Наука. KAZLAUSKAS, J. 1967: [Rec.:] Acta Baltico-Slavica III, Baltica in honorem Iohannis Otręb ski, Białystok, 1966. Baltistica 3 (2), 237–243. (= Kazlauskas 2000: 236–246.) KAZLAUSKAS, J. 1968: [Rec.:] Lietuvių kalba tarybiniais metais, Vilnius, 1967. Baltistica 4 (2), 322–327. (= Kazlauskas 2000: 279–286.) KAZLAUSKAS, J. 2000: Rinktiniai raštai 2, sudarė A. Rosinas. Vilnius: Mokslo ir enciklopedijų leidybos institutas. KRYLOV, S. A. 2002: «Русское именное словоизменение» А. А. Зализняка 30 лет спустя: Опыт ретроспективной рецензии с позиций неоструктуралистской морфологии. Zaliznjak 2002: 699–748. LAIGONAITĖ, A. 1978: Lietuvių kalbos akcentologija, Vilnius: Mokslas. LEVIN, J. 1994: Sign of Jod (The Semiotic Function of /j/ in Lithuanian). Baltistica, 4 priedas, 82–89. MIKALAUSKAITĖ, E. 1975: Lietuvių kalbos fonetikos darbai, Vilnius: Mokslas. PAKERYS, A. 1995: Lietuvių bendrinės kalbos fonetika, Vilnius: Žara. PAULAUSKIENĖ, A. 1983: Lietuvių kalbos morfologijos apybraiža, Kaunas: Šviesa. PAULAUSKIENĖ, A. 1989: Gramatinės lietuvių kalbos vardažodžių kategorijos, Vilnius: Mokslas. PAULAUSKIENĖ, A. 1994: Lietuvių kalbos morfologija. Paskaitos lituanistams, Vilnius: Mokslo ir enciklopedijų leidykla. PUPKIS, A. 1966: Lietuvių kalbos afrikatų ir atitinkamų priebalsių junginių sudėties klausimu. Kalbotyra 14, 107–117. PASTABOS 129 Stepanov, Ju. S. 1972: Ударение и метатония в литовском глаголе. Baltistica, I priedas, 169–183. Stepanov, Ju. S. 1997 a: Непарадигматические передвижения ударения в индоевропейском. (Исходные положения.) Вестник Московского университета. Серия 9. Филология № 5, 63–80. Stepanov, Ju. S. 1997 b: Непарадигматические передвижения ударения в индоевропейском. (I. Вокруг законов Ваккернагеля и Лескина.) Вопросы языкознания, 1997, № 4, 5–26. Stundžia, B. 1980: Vadinamosios bendrosios giminės vieta lietuvių kalbos morfologinėje sistemoje. Kalbotyra 31 (1), 79–87. Trubeckoj, N. S. 1960: Основы фонологии, Москва: Издательство иностранной литературы. Trubeckoj, N. S. 2000: Основы фонологии, Москва: Аспект Пресс. Ulvy das, K., red. 1965: Lietuvių kalbos grammatika 1. Fonetika ir morfologija, Vilnius: Mintis. Vaitkevičiūtė, V. 1961: Lietuvių literatūrinės kalbos balsinės ir dvibalsinės fonemos. Lietuvių kalbotyros klausimai 4, 19–47. Valeckienė, A. 1998: Funkcinė lietuvių kalbos grammatika, Vilnius: Mokslo ir enciklopedijų leidybos institutas. Zaliznjak, A. A. 1967: Русское именное словоизменение, Москва: Наука. (= Zaliznjak 2002: 1–370.) Zaliznjak, A. A. 2002: «Русское именное словоизменение» с приложением избранных работ по современному русскому языку и общему языкознанию, Москва: Языки славянской культуры. Zinkevicius, Z. 1966: Lietuvių dialektologija, Vilnius: Mintis. Aleksejus Andronovas St. Petersburg State University Department of General Linguistics Universitetskaya nab. 11, 199034 St. Petersburg, Russia [email protected] Gauta 2002 12 05 NESUSIPRATIMAI DĖL DAŽNINIO ŽODYNO Acta Linguistica Lithuanica XLVI tome Artūras Judžentis, apžvelgdamas Vito Labučio mokslinę veiklą (p. 230–248), atpasakoja kai kuriuos teiginius iš jo recenzijos apie mudviejų žodynus – Dažninių dabartinės rašomosios lietuvių kalbos žodyną (mažėjančio dažnio tvarka) (1997) ir kitą žodyną tuo pačiu pavadinimu, tik išdėstytą abėcėlės tvarka (1998) (tiesa, supainiojęs mūsų vardus ir pavardes, žr. p. 245). Panūdome autoriui viešai replik(u)oti dėl dėstomų ne jo, o Labučio teiginių todėl, kad, Judžentis nepasinaudojus dalyvinės kalbos galimybėmis, telieka daryti prielaidą, jog tiek apžvalgos autoriaus, tiek Labučio nuomonės sutampa: esą dažninius žodynus galima rengti nesilaikant lingvostatistikos reikalavimų (plg. Labučio teiginį: „Remiantis statistikos principais pagrįstomis tekstų imtimis ir įsitikinus, kad pro jų tinklą nuolat išslysta koks ungurys ar žiobris, tekstų imtis turbūt reikėtų koreguoti“. (Kalbos kultūra 72, 1999, p. 113.)) Judžentis, be jokių komentarų pakartojęs ne vien Labučio, beje, kaip ir kitų, nepakankamai įsigilinusių į lingvostatistinių žodžių sąrašų ir žodynų sudarymo principus, priekaištą dėl kurių nors žodžių nebuvimo tame sąraše (… „pasigendama (…) buitinių realijų, augalų bei gyvūnų pavadinimų, kultūros srities ir su intelektine veikla susijusių žodžių…“), ir pakėlino priminti, kad dažniniai žodynai negali būti papildomi vienais ar kitais, kad ir labai pageidaujamais, žodžiais. Tiesiog tokia lingvostatistinių metodų prigimtis: nei pridėti, nei išmesti nieko negalima – statistiškai aprašytinas realiai egzistuojantis