Referential Expression Type and Definiteness as Factors in Asymmetrical Object Marking in Modern Hebrew

Alexeeva Mariya Senior Lecturer, Department of Semitic and Hebrew Studies, St Petersburg University 199034, Russia, Saint Petersburg, Universitetskaya Nab., 7-9 maria-alexeeva@inbox.ru Other publications by this author

Problem statement. Asymmetric object labeling describes a phenomenon in which the second argument of a divalent verb is realized in the grammatical role of a direct object, but the morpho-syntactic encoding of this argument varies between the zero and the marked exponent of the accusative. The problem of asymmetric object marking in modern scientific discourse is usually considered in terms of "differentiated object marking" ^{[20, 21]}, which, in turn, is a special case of the broader phenomenon of differentiated argument marking ^[18]. The term "differentiated marking of an object" ("differentiated object marking", or "variable design of a direct complement" ^[5]) was introduced by G. Bossong in 1985, and even then the variability of object marking was recorded in more than 300 languages of the world ^[9].

Today, there are several basic approaches in the scientific literature that explain the variability of object labeling ^{[3, 4, 10, 19]}. The most influential to date is the theory of optimality, presented in the work of J. Eissen ^[6]. It is in this work that the fundamental importance of two factors-the licensors of differentiated object labeling - is postulated: definiteness and animateness, and modern Hebrew is given as an example of the influence of the determinateness factor ^{[6, p. 436]}.

Indeed, modern Hebrew is traditionally perceived as an example of a language in which there is asymmetric object labeling and which, within the framework of asymmetric object labeling, uses the marker of the accusative ?et, encoding the O-participant in a transitive clause differently depending on the status of the definiteness of the nominal group (IG) that encodes it. An important feature of the perception of the definite/indefinite status of IG in modern Hebrew is that this category is binary, and its interpretation is based on the presence/absence of formal indicators of determination: a definite article, possessive suffixes, proper names.

Research papers on the case system of modern Hebrew [for example, 11, 13, 16, 22], for the most part, they are performed within the framework of the theoretical paradigm proposed by N. Chomsky's generative grammar, and appeal to grammatical descriptions and native speakers' own intuition. So far, no systematic attempts have been made to use the theoretical basis of functionalism to analyze object labeling in modern Hebrew, as well as to apply corpus linguistics methods for such analysis. The only work combining both was published by P. Beckins and is based on the material of the Hebrew language ^[7]. Thus, this research is the first study of asymmetric object labeling in the modern Hebrew language from the point of view of a functional approach using corpus analysis methods, which is its scientific novelty.

At this stage, one of the main features of functional linguistics is the lack of a unified global theory that provides a certain view on all aspects of language. However, research within the framework of functionalism is conducted by numerous scientists around the world both at the general theoretical level and on the material of a variety of languages and dialects ^{[1, 2, 12, 14, 17]}. The relevance of the research, therefore, lies, on the one hand, in the application of a functional approach, and on the other, in addressing the topic of differentiated case marking, to which numerous works of recent decades have been devoted.

The purpose and methods of the study. The purpose of this study is to show that the correlation between the presence of formal indicators of determination in IG and the presence of an accusative marker in the design of a direct object in modern Hebrew is not absolute. To do this, we will consider the ways of expressing the category of "certainty" in modern Hebrew in different nominal groups, which in modern Hebrew can encode the O-participant in a transitive clause, which are the subject of the study, and analyze the correlation between the status of certainty/uncertainty of the IG and the way it is formed in the position of a direct object.

To obtain objective and statistically substantiated information about the variability of asymmetric object labeling in the speech of native Hebrew speakers, we will use the methods of quantitative and comparative analysis developed by corpus linguistics on the material of two research corpora Hebrew Objects General Corpus, with a volume of about 52,000 word uses, and Hebrew Objects Targeted Corpus (HOT corpus), with a volume of about 49,000 word uses, compiled on the basis of the online corpus of modern Hebrew Hebrew Web 2021 (heTenTen21), hosted on the SketchEngine platform ^[24].

The main types of nominal groups in modern Hebrew and indicators of determinationAccording to the classification of Sh .

Wintner, IG can include various elements, the order of which is strictly fixed [Wintner]. In particular, quantifiers (shlosha ("three")), kol ("every/every/all/all"), kama ("several")), determinants (oto "the same/the same") and the proper definite article (h a-) occupy a position ahead of the vertex, and in the postposition to the vertex, the following can be used: IG defining the vertex, adjectives and ordinal numerals, demonstrative pronouns, possessives (for example, sheli ("my"), shel dan ("Dana")), prepositional groups and relative subordinate clauses. Formalized by the indicator of determination (including the article) The IG is considered certain. An indefinite IG cannot be formed by indicators of determination (there is no indefinite article).

The marker of the accusative ’et may be omitted before a certain IG in some types of texts requiring brevity, for example, in press headlines, especially if the direct addition does not follow immediately after the preposition ^{[15, p. 158]}.

Assuming a potential differentiation by the definiteness parameter of the above-mentioned structurally different IGS, we divided them into four categories (proper names, defined IGS with other indicators of definiteness, indefinite IGS, pronouns) and conducted a quantitative and comparative analysis of asymmetric object labeling for each category.

In the vast majority of cases (with the exception of members of the same family, designated by the plural form, and special forms when naming holidays) proper names in modern Hebrew do not accept indicators of determination, but demonstrate obligatory object labeling regardless of the presence of the article (1, 2) or restrictive definitions (3):

Certain IGS, formed by other indicators, exhibit significantly greater variability. According to the HOG corpus, IGS consisting only of a noun and decorated with a definite article or possessive suffix are marked in 100% of cases (286/286 and 60/60, respectively), as well as more complex IGS, decorated with the article h a-: IG with an indicative pronoun (8/8), with an ordinal numeral (1/1), with a definite determinative (4/4). Isolated cases of deviation from the above principle are recorded by the HOG corpus for a certain IG with an inconsistent definition (conjugate state construction) (96/97) and a certain IG with a prepositional genitive structure (with the preposition shel) (42/43). A significant degree of variability of asymmetric object labeling or a complete ban on it is recorded in two types of IG in the category of "certain": IG with an indicative pronoun, not decorated with a definite article (IG+ EdicT (art.-)) (0/1) (4), and IG with the quantifier kol "all, all, each" (79% marked, 19/24) (5).

  The variability of the object labeling of the IG with the kol quantifier is rightly explained by the difference in the design of the IG with different meanings ("all objects of the set" and "the whole object" vs "each/no object of the set"), however, the variability of the object labeling of referential expressions that do not accept the indicators of determination (see, in particular, example (5)) up to it has not yet been considered in the scientific literature and, as far as we know, has not even been described.

Due to the low frequency of some types of referential expressions that make up the subject of the study, to determine the statistical significance of the results obtained in the HOG corpus, a second HOT research corpus was formed, which recorded only those types of IGS that demonstrated deviations in the HOG corpus. According to the HOT case, IG type IG +EdicT (art.-), contrary to the generally accepted opinion, demonstrate the optionality of labeling (2% marked, 5/222). Note that all contexts in which this IG is framed with the marker 'et(5/5) are represented by IG with the pronoun zo (FS) "this", which indicates the presence of an additional factor, unaccounted for by previous studies, licensing object labeling in these contexts.  The optionality of labeling is also confirmed in the case of the kol quantifier, especially in examples like (5), where the O-participant of the transitive clause encodes the relative pronoun ma, accompanied by a determinative subordinate: 55% is labeled (9/20).

Undefined IGS encoding the O-participant in a transitive clause must demonstrate a prohibition on labeling. According to the data of the HOG corpus, this principle is valid for nouns unformed by the article (0/471), indefinite genitive constructions with prepositional formalization (0/12), indefinite IGS with determinants having the semantics of uncertainty (0/6) and with indefinite quantitative numerals (0/11). However, constructions denoting the partitive formally tend to at the indefinite end of the spectrum, they demonstrate a high degree of optionality of object labeling (75% are labeled, 25% are not, according to the HOG corpus), for example:

The referential expression "one of ...", in particular, demonstrates the probability of object labeling at 78% (57/73).

The phenomenon of optional partitive labeling in question has been confirmed experimentally ^[16], but at the moment it has not found an acceptable explanation (isolated attempts made within the framework of a formal approach to syntax either ignored this phenomenon ^[11], or did not consider the discursive-pragmatic features of this type of referential expressions ^[16] which, in our opinion, underlie the variability of object labeling in modern Hebrew.

As for pronouns (in this case we are not considering actualizers clarifying the reference of the noun ("this table"), but independently coding the referent), then, according to the theory of optimality, pronouns are located in the leftmost position on the Scale of certainty ^[6], therefore they must be subject to mandatory labeling, because their referent is uniquely identified communicants.

However, this fact is not confirmed by the corpus analysis data.  According to the HOG corpus, less than half of the referents encoded using demonstrative pronouns (16/33. 48%) are marked, which should unambiguously indicate the referent (7).

Interrogative pronouns are more naturally associated with uncertainty than with certainty, due to the fact that the speaker does not know which referent is meant. As a confirmation of this, in some languages, interrogative words also act as indefinite pronouns, since both assume an "information gap" ^{[8, p. 226]}. However, when encoding a direct object, interrogative pronouns in Hebrew, as well as demonstratives, are marked differentially depending on the form: ma "what" is typically not marked, but demonstrates some optional labeling in the corpus HOT (1/62), mi "who" is marked obligatorily (30/30). This phenomenon is noted by researchers both in Hebrew and in modern Hebrew ^[15], but finds an explanation only within the framework of a discursive-pragmatic approach ^[7].

The situation is more complicated with the relative pronouns mi "(one) who" and ma "(that) what", encoding the O-participant in the transitive clause, the differentiated object labeling of which has not been considered in scientific works at the moment. In this case, not only is it not possible to explain the variability of asymmetric object labeling by a certain status of the referential expression, but also the correlation with the interrogative pronouns mi and ma, which have a sufficiently strict differentiation of object labeling, is not obvious.

Pronouns of other categories demonstrate the expected design in the position of a direct object: mutual and object are marked obligatorily (5/5 and 122/122, respectively), indefinite pronouns are not marked (0/7).

Conclusions and conclusion. The analysis of the referential expressions encoding the O-participant in the transitive clause, carried out in this study on the basis of corpus data on the material of modern Hebrew, allows us to conclude that, despite the unconditional existence of a correlation between the category of "certainty" and the way of registration of the IG in the position of a direct object, full consistency with the research and didactic with the principle "everything definite is marked, everything indefinite is not", only one category of the four considered, proper names, is marked obligatorily, demonstrates in speech. The remaining categories demonstrate the optionality of object labeling within one or more types of referential expressions within each group. In the category of "certain IGS", the optionality ranges from 2% for IGS with demonstrative pronouns not decorated with a definite article, to 55% for IGS with the quantifier kol. Formally indeterminate IGS having the meaning of incomplete coverage (partitive constructions), on the contrary, demonstrate a very high probability of object labeling, 75%, especially for referents encoded using the quantitative numeral ?ehad/?ahat "one". Pronouns are a very heterogeneous class, which, nevertheless, frequently demonstrates the optionality of labeling: demonstrative pronouns, whose referent must be uniquely identified, are marked in less than half of the cases, interrogative ones are strictly differentiated by form, while relative ones show a high degree of variability (66% and 85% are marked depending on the form), for which no explanation has yet been offered in the literature.

Thus, our study clearly demonstrates that a certain status of the IG encoding the O-participant in the transitive clause cannot be considered the only factor licensing asymmetric object labeling in modern Hebrew. To determine additional factors, additional research is required, based primarily on data on the real speech of native speakers, allowing to operate not only with formal characteristics of IG, but also to take into account the discursive and pragmatic features of referent-coding expressions, such as referential status, animateness, activity in discourse and the degree of topicality.