ACTA LINGUISTICA LITHUANICA

LII (2005), 157–166

REMARKS

Remarks

ON THE CLASSIFICATION OF SYNTACTIC RELATIONS

The ideas presented here were prompted by an interesting review by Prof. Vytautas Ambrazas (ALL 52, 167–170) of the 2003 publication Studies in Syntactic Relations. This review, which contains valuable polemical remarks, especially concerning the theory of syntactic relations presented in the publication mentioned above, shows that the authors’ aim—to encourage theoretical discussion of issues in Lithuanian grammar—has been achieved. It is particularly gratifying that the discussion has been initiated by a distinguished authority on the synchronic and diachronic syntax of Lithuanian, Prof. Vytautas Ambrazas, author of many works that have become classics. The professor himself has repeatedly taken part in seminars held by the Grammar Department of the Institute of the Lithuanian Language, observing and encouraging the crystallization of ideas as a sympathetic yet perceptive and critical listener. It is to be hoped that an ever wider circle of researchers will join the discussion.

No syntactic theory is perfect, and the theory of syntactic relations presented in Studies in Syntactic Relations is no exception. For convenience, I will refer to it below by the abbreviation “SRT theory.” Prof. Ambrazas rightly draws attention to some problems associated with it. These problems are difficult to refute; the worst course would be to sweep them under the rug. However, the authors of the SRT conception decided to accept them according to the principle of the “lesser evil.” I would therefore like to draw attention here to some greater shortcomings that SRT theory helps avoid.

The reviewer observes that the concepts of government and modification used by the SRT authors are very broad. For example, modifiers are taken to include attributes, adverbials, and predicative attributes (this needs clarification: under SRT theory, only non-obligatory verbal free predicatives are considered modifiers). It is difficult to deny that this observation is partly justified: the “double syntactic relation” is rejected, under which DLKG distinguishes nominal parts of the predicate and predicative attributes. The problem is that this double syntactic relation leads to contradictions that are difficult to resolve.

A predicative attribute is defined by two features: it attaches to a verb and agrees with the subject (Broliai grįžo linksmi) or the direct object (Visada matau Joną linksmą). This is what distinguishes a predicative attribute from an adverbial in a sentence such as, for example, Nudažiau duris žaliai. In this sentence, the adverb žaliai does not agree with anything, so in this case the syntactic relation is single. But why could one not claim that the adverb žaliai attaches to the noun duris just as it does to the verb? After all, semantically, the property it denotes is assigned to the door, not to the painting: Nudažiau duris žaliai means ‘by painting, I made the door green’. Such a claim cannot be proved, but neither can it be disproved. For attachment is a “syntactic relation” that cannot be seen—it is not manifested by morphological features. Thus, the claim that in the sentence Nudažiau duris žaliai the word žalias is linked by a double syntactic relation is just as valid as the claim that it is linked by a single syntactic relation

158 |

PASTABOS

relation. That one and the same claim and its opposite can be equally valid is, of course, undesirable.

At this point, a clarification is needed: we see nothing wrong with distinguishing a type of syntactic relation that has no morphological features. In the SRT conception, syntactic relations are differentiated without any regard to morphological features. The difficulty lies in placing purely morphological phenomena on the same level as syntactic dependency relations. A morphological dependency relation (in which the inflectional form of one word is determined by the morphological features of another) can conventionally be called a type of syntactic relation, but it is forgotten that the heuristic procedures are entirely different depending on whether the phenomenon in question is morphological—agreement—or a phenomenon without morphological features—attachment. In the case of agreement, a morphological connection is established, and from it the existence of a syntactic relation is inferred: it is precisely on this basis that both mutual connection and a double syntactic relation are defined. If there were no agreement, the subject could be understood as governed by the verb, and the predicative attribute as attached to the verb. In the case of attachment, the procedure is the reverse: one begins with the assumption that there must be a syntactic dependency relation between two words, and, since this relation can be neither government nor agreement, it is assumed to be attachment. For example, in the sentence Jis greitai atbėgo, it is assumed that the word greitai depends on the verb, and this dependency relation is called attachment because it has no morphological marker.

We do not wish to reject this assumption in any way. In the sentence mentioned, the word greitai must depend on something, and the assumption that it depends specifically on the verb atbėgo seems entirely plausible. This assumption is also justified in the case of the sentence Nudažiau duris žaliai. It is likewise justified when analyzing the sentence Visada matau Joną linksmą. The adjective linksmas here depends on the verb matau, and that is already enough to assign this adjective a place in the hierarchical syntactic structure of the sentence. It is entirely unnecessary to make the additional assumption that this adjective also depends on the noun Joną. It is true that its number and gender form depends on the number and gender of the word Joną, but such morphological dependency can link even words occurring in entirely different sentences (e.g., Ant stalo gulėjo pinigai. Išs idėjau juos į kišenę), and in such a case no one sees any need to establish a syntactic dependency relation between them. The worst consequence is that, once a syntactic relation between the words Joną and linksmą in the sentence mačiau Joną linksmą is postulated on the basis of agreement, it is natural to assume that a syntactic relation also exists between the words duris and žaliai in the sentence nudažiau duris žaliai. After all, everything is similar here both semantically and syntactically; the only difference is the difference in parts of speech: either an inflected adjective or an uninflected adverb is used. But can such a difference be syntactic? Probably not; therefore, once we accept the view that syntactic dependency can be double, that is, that one word can depend on several other words, it is natural to assume that the words duris and žaliai are linked by an attachment relation. One might respond that in this case (unlike, for example, in the sentence Jis greitai atbėgo) there is no a priori reason to suspect that a syntactic relation exists. But then there is no such reason in the case of the construction Mačiau Joną linksmą either. In this construction we see only a morphological connection, which—as has already been noted—does not entail a syntactic relation.

As can be seen, the logical contradictions in the traditional model of syntactic relations arise from treating purely syntactic phenomena (attachment) and purely morphological phenomena (agreement) at the same level. In principle, there is no reason to include agreement in the classification scheme for syntactic relations. Explicitis verbis, Prof. Ambrazas does not propose this in his review; however, the proposed notions of mutual connection and double syntactic relation rest on the erroneous assumption that agreement is a syntactic relation,

PASTABOS

| 159

or presupposes a syntactic relation. Once this assumption is abandoned, there is no need to operate with the notions of mutual connection and a twofold syntactic relation, and then only two notions remain from the traditional scheme of syntactic relations: government and adjunction. These notions should be defined precisely as they are in the SRT conception.

In Prof. Ambrazas’s view, applying the notion of government only to unilateral subordination (while leaving mutual connection and a twofold relation as separate types of relation) “would make it possible to meet the need, acknowledged by the authors themselves, to define sentence-unit functions on the basis of syntactic relations.” In fact, the ability of both syntactic-relation models—the one proposed by Prof. Ambrazas and the one proposed by the SRT authors—to distinguish sentence constituents according to the nature of the syntactic relation is roughly the same. The SRT conception does not distinguish subjects, objects, and obligatory predicatives on the basis of syntactic relations. However, it does distinguish objects from adverbials and obligatory predicatives from free ones, which the model set out in DLKG cannot do. If we adopt the improved version proposed in Prof. Ambrazas’s review, in which government and adjunction are no longer defined according to morphological criteria, it does indeed become possible to distinguish objects (governed) from adverbials (adjoined) on the basis of syntactic relations. This cannot be done in the case of the “nominal part of the predicate” and “predicative modifiers.” It is claimed that in the sentence Jonas yra linksmas, the word linksmas is not a separate sentence constituent, because the copula būti has no independent meaning. “Predicative modifier” is recognized as a sentence constituent in the sentence Jonas grįžo linksmas. But what function does the word linksmas have in the sentence Jonas atrodo linksmas? DLKG gives two different answers to this question: on pp. 507–508, the verb atrodyti is listed among non-pure copulas, whereas on p. 589 the sentence Mokytojas atrodė piktas is given as an example of a predicative modifier. Of these solutions, the second should be considered better, since it is difficult to claim that the verb atrodyti has no meaning, in the same way as būti. In the SRT conception, the adjective linksmas in the sentences Jis atrodė linksmas and Jis grįžo linksmas is linked by different syntactic relations—in one case, government, and in the other, modification. This explains why the adjective is obligatory in the first case (*Jis atrodė is an incomplete sentence), but not in the second. In essence, this difference is entirely analogous to the difference between objects and adverbials, which Prof. Ambrazas himself tends to explain as a difference in syntactic relations. But in the case of “predicative modifiers,” this difference is concealed under the same label, “twofold syntactic relation.” Attention was drawn earlier to the fact that it would be better to abandon the twofold syntactic relation, since it leads to fundamental contradictions. Like the notion of a twofold syntactic relation, the notion of mutual connection is also flawed. Therefore, one should doubt whether, on the basis of the conception presented in DLKG, it is possible to distinguish subjects from objects by syntactic relation. Only generative grammar, which relies on immediate-constituent analysis and divides a sentence into a subject and a verb phrase at the very first stage of parsing, succeeds in doing this convincingly. The SRT authors do not apply this method, and therefore acknowledge that the distinction between subjects and objects is not a difference in types of syntactic relation. Their only syntactic difference concerns their position in the hierarchy of noun phrases (English obliqueness hierarchy)—other differences are morphosyntactic.

In Prof. Ambrazas’s view, the notion of mutual connection would help avoid the “contradictory term ‘obligatory adverbials.’” This term is used in the SRT conception to describe, for example, prie ežero in the sentence Namas yra prie ežero. It must be admitted that the term itself is not very successful, but it can be replaced with a better one. A term is one thing; the essence of the matter is another. The point is that the “obligatory adverbial” case is intermediate between government and modification. The sentence is locative, and the noun phrase expressing location undoubtedly occupies an argument position. Occupying an argument position is generally characteristic of objects, but calling prie ežero an object is prevented by a certain peculiarity of its expression: strictly speaking, the predicate should be formulated as ‘be at’, and this predicate opens an argument

160 |

PASTABOS

positions for the names of a thing and a location. But the spatial relation is expressed not by the verb, but by the preposition used with the noun phrase. Thus ežero resembles an adverbial in meaning and form of expression (or, to put it in more generally accepted terms, a modifier)—except that in this case the modifier expresses not an additional, but the main predication of the sentence. More precisely, it is not the noun ežeras itself that expresses the predication, but the preposition prie. But the usual rules of syntactic parsing do not allow the sentence to be parsed as Namas / yra prie / ežero. The accepted parse is instead Namas / yra / prie ežero. Such an argument, syntactically realized as a modifier, is called an obligatory adverbial according to the principle “call it what you will; it won’t make any difference.” The fact that there may be intermediate cases between government and modification is not contrary to the theory. The boundaries of linguistic categories are generally fuzzy; what matters most is that the tools used by the researcher are sharp. And the alternatives to “obligatory adverbials” are not much better. To claim that in the sentence Namas yra prie ežero the sequence of words prie ežero is not a sentence constituent, but only part of a sentence constituent, is simply to evade an answer to a difficult question. How can we explain that the sequence of words prie ežero, which in one sentence (Namas yra prie ežero) is “nothing,” becomes a sentence constituent in another sentence (Jis pasistatė namą prie ežero)?

One could answer this question by saying that in the sentence Namas yra prie ežero, prie ežero is in fact the true predicate: this would be supported by the possibility of omitting the present-tense form of the verb būti, a phenomenon to which Prof. Ambrazas attaches great importance. He comments on the SRT authors’ proposal to treat the obligatory predicative with the verb būti as governed as follows: “Thus, it will be necessary to assume that in the sentence Knyga įdomi the zero copula (i.e. a marker of tense and mood without lexical meaning) governs both nominatives as obligatory complements to the meaning of the verb.” But if the zero copula is a problem for the SRT conception, it also poses problems for the DLKG and Prof. Ambrazas’s conception. If the zero copula cannot govern, then it cannot enter into a mutual connection either—in other words, the subject of the sentence Knyga įdomi cannot be defined syntactically. As for the meaning of the copula, it is important to understand that whether or not we assign it lexical meaning is a matter of convention, determined by considerations of theoretical consistency. The copula būti does not always connect referential expressions with predicative words (as in the example Jonas yra linksmas); sometimes it also connects referential expressions with other referential expressions, for example, in the sentence Jonas yra mūsų mokytojas, where mūsų mokytojas is not a predicate, but a definite description (English definite description) that denotes a person. In this case, the copula yra expresses the identity of the referents of two expressions (‘The man named Jonas and the man who teaches us are one and the same person’). The assumption that the copula has no meaning does not yield good results here, because a sentence cannot consist solely of referential expressions and lack predicative content. In other words, at least in some cases the copula must be assigned lexical meaning. In any case, lexical meaning has to be assigned to “non-pure copulas” (DLKG 507–508), such as atrodyti, apsimesti and the like, whose complements in the DLKG conception are left in no-man’s-land somewhere between the “nominal part of the predicate” and the “predicative modifier.” If we acknowledge that būti, atrod yti also have a certain lexical meaning, their complements can be described uniformly as governed elements. This introduces a certain redundancy into the description of sentences such as Jonas yra linksmas, since we must acknowledge that the word linksmas itself already has predicative meaning. But of the various evils, redundancy is probably the least, since no one would deny that redundancy occurs in language. Semantically, the copula in the sentence Jonas yra linksmas is, of course, redundant. But a syntactic description that pretends it is not there gives rise to far worse contradictions.

PASTABOS

161

The search must continue, and the confrontation of ideas and concepts will undoubtedly one day make it possible to formulate a better theory of syntactic relations than the one formulated by SRT.

Axel Holvoet Lithuanian Language Institute P. Vileišio St. 5, 10308 Vilnius, Lithuania

Received 2005 05 13

TRIVIA 7

We have long known and been taught that the allophone ŋ of the sonorant n is soft only before k’ (this was established by Prof. Aleksas Girdenis (Kalbotyra 48–49 (1), pp. 165–167), and presented by the author of these lines in the Dictionary of Lithuanian Dialects (2001); this is how it is pronounced throughout Lithuania (some southern Aukštaitians, called Dzūkians, replace it with n under Slavic influence).

Dabar žurnale Lituanistica (2004, 4 (20)) pasirodė (po Vilniaus pedagoginio universiteto mokslininkų darbų) Arimanto Raškinio, Gailiaus Raškinio, Astos Kazlauskienės straipsnis „Lietuvių bendrinės šnekamosios kalbos garsų fondo kūrimo principai“ (p. 53–62). Jame teigiama, turbūt pagal prof. Antaną Pakerį (kaip rodo prof. A. Girdenio minimas straipsnis), kad bendrinėje kalboje turi būti tariama an’ks’t’i, tan’km’ė, án’kš’t’is (p. 56), o paprastai visur lietuvių tariama ankš’t’i, tankm’ė, á.nkš’t’is ir pan. Tai neįtikinamas siūlymas ir neigiamas dalykas (gal tame straipsnyje nereikėjo keisti ir daiktavardžių su priesaga -ybė priegaidės – tikimybė (p. 61) = tikimybė).

Injustice also comes from VPU to LKI through its students: the important publication Anthology of Lithuanian Dialects (2004) gives luŋ’.kt’ “to bend” (Širvintiškiai) (p. 162), an’kš’č’áu (Pákražantis) (p. 215), deŋ’.kt’ (p. 18—the theoretical section). It even goes against the book’s consultant, Prof. Habil. Dr. A. Girdenis! There is even bě’kt’ (p. 62). And this is from the Šiauliai dialect...

TRIVIA 8

In northern Žemaitija (all Dounininkai, northern Dūnininkai, called Varniškiai, and northern Donininkai—from Saūgai northward), stress is retracted to the beginning of the word from all short and circumflex endings (universal stress retraction). In those dialects, stress may also fall on the preposition of an inflected word (as Jonas Jablonskis observed while editing Antanas Juška’s dictionary). In the third volume of Questions of Lithuanian Linguistics (1960) (pp. 62–63), this is described for Žemaitian—in the Kuršėnai, Papilė, Kaunatava, Raudėnai, and other dialects; similar phenomena have been reported in the Kalnėlis dialect (P. Kniūkšta), Laukuva (V. Grinaveckis), and even the western Aukštaitian dialect of Šakyna (A. Jonaiytė); I myself came across the pronunciation vā·re aŋ darbu· near Šiupyliai, not far from Gruzdžiai. Many dialectologists believed this, especially Academician Zigmas Zinkevičius and the authors of the first Anthology of Lithuanian Dialects.

The dialectologist Aleksas Girdenis and his student Dr. Lina Murinienė now discuss this issue at length in the article “The Specific Features of Stress Retraction onto a Proclitic in the Eastern Northern Žemaitian Dialect” (Baltistica 36 (1), 2001, pp. 87–102). Having revisited the problem raised since 1960, the article examines it very persuasively, with examples deciphered from various recordings, and even indicates its frequency,