Animacy and Referential Status as Factors of Asymmetric Object Marking in Modern Hebrew (based on Interrogative and Relative Pronouns)

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

Introduction. The phenomenon of differentiated marking of actants (Differential argument marking) is characterized by the presence in different structured languages of regularly occurring cases of deviation from the prescribed method of encoding nominal groups acting as a subject, object, possessor or locative, under the influence of a certain number of factors-licensors. The regularity of alternating coding methods under the influence of the same factors in languages with different structures allows researchers to talk about the existence of unified mechanisms that license the differentiated labeling of various actants. This phenomenon is of great interest to researchers [for example, 1-3, 14, 15, 19] and generates a number of more specific issues related to the differentiated marking of the subject (Differential subject marking), locative (Differential locative marking) or object. Differentiated object marking (the term was introduced by G. Bossong in 1985 ^[9]), in turn, is divided into symmetric and asymmetric, describing two ways of morpho-syntactic encoding of the second argument of a divalent verb, which is implemented in the grammatical role of a direct object: symmetric labeling is implemented by alternating accusative and non-causative markers, and the asymmetric — zero and non-zero accusative markers.

We can talk about several basic approaches that explain the variability of object labeling ^{[5, 7, 10, 15, 20]}. The first in this series and still recognized as one of the most influential is the theory of optimality, presented in the work of J. Eissen ^[7]. In this work, two factors were named-the licensor of differentiated object labeling, which formed the basis of subsequent hypotheses: the factors of certainty and animateness. In the future, these factors, one way or another, were present among the motivators regulating object labeling, and in a general way it is possible to formulate a list of licensor factors mentioned in various works as follows: referential status, animateness, topicality, values of species-specific and action categories, polarity, inclination ^[19].

The first two categories in this list, along with the category of certainty, which some authors prefer to consider separately, correlate with the so-called "degree of individualization of the object". For example, according to P. Hopper and S. Thompson, a high degree of individualization correlates with the animateness, definiteness and concrete-referential status of an object, including expressed by proper names and singular forms ^{[16, p. 253]}, and according to J. According to Khan, an individualized object is most likely to correspond to the following parameters: the certainty and/or presence of a quantifier (qualifier) as part of a nominal group, the referency of the nominal group and the animateness of the referent, its encoding with a proper name, as well as a high degree of activity of the referent in discourse ^{[18, pp. 469-470]}.

The purpose of this work is to identify the correlation between the variability of asymmetric object labeling in modern Hebrew and two of the listed factors-licensors, namely "referential status" and "animateness". To achieve this goal, the interrogative and relative pronouns encoding the O-participant in the transitive clause in the Hebrew language will be considered, which will form the subject of the study and, among other things, justify its scientific novelty, since this type of referential expressions encoding the patient participant of the situation in modern Hebrew has not been the subject of scientific discussion until now.

Due to the relatively low frequency of this type of referential expressions in natural speech, the author decided to resort to methods developed by corpus linguistics, as a result of which an experimental corpus of Hebrew Objects Targeted Corpus (HOT corpus) was formed for quantitative and comparative analysis, 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 ^[22]. To conduct a comparative analysis in the corpus, in addition to the actual interrogative and relative pronouns, other types of referential expressions demonstrating optional object labeling were also marked up according to the above parameters: nominal groups with the kol quantifier "all/everyone/everyone/everything" (62% marked (197/316)) and with demonstrative pronouns that are not decorated with a definite article (3% (6/223)), constructions of the partitive (38% (60/156)) and demonstrative pronouns that independently encode the O-participant in the transitive clause (48% (139/290)). A cursory calculation confirmed the hypothesis of the presence of asymmetric object labeling also for the type of referential expressions under consideration: interrogative pronouns in the HOT corpus were labeled in 34% of the presented contexts (31/92), and relative pronouns in 78% (100/129). 

The "animateness of the referent" factor

At the first stage, all referential expressions encoding the O-participant in the transitive clause were marked by the "animateness" parameter, which made it possible to draw a preliminary, but very indicative conclusion about the correlation between the animateness of the referents and the way of labeling the referential expressions encoding them (Table 1) (only data on referential expressions were excluded from the calculations, referring to propositional antecedents, due to their lack of the parameter "animateness/inanimateness"). To understand the degree of significance of the correlation of the proposed parameters, the parameter "p-value" (significance level) was calculated here and further, which reflects the result of the chi-square test, used in statistical studies to determine the probability that this correlation is random.  A p-value of less than 0.05 indicates that the probability of a random correlation between variables is less than 5%, which is usually considered statistically significant.

The p-value value for the parameters in Table 1 is < .00001, i.e. the probability that the correlation between the parameters "animateness" and "design of a referential expression using an accusative marker 'et' in the contexts fixed by the HOT corpus is random is less than 1%.

Table 1.

However, it is difficult to determine to what extent the animateness factor of the referent licenses the asymmetric object labeling of formally defined indicators of the determination of nominal groups encoding this referent (for example, kol ha-sfarim all ART.books "all books"), because the "definiteness of the nominal group" itself is recognized as a sufficient basis for its labeling in positions of a direct object ^{[13, 21]}. 

However, the animateness parameter can become the main or one of the main licensing factors for those referential expressions that do not have a clearly defined status of certainty. The influence of this factor will be especially indicative on the example of interrogative and relative pronouns, which we have set out to consider in this paper.

Interrogative pronouns

Both in modern and, for example, in Hebrew ^[8], there are two interrogative pronouns encoding the patient participant of the situation: the pronoun m i "who" encodes the animate referent, whereas m a "what" is inanimate. In this case, the pronoun mi is typically formed with an accusative marker (1), whereas ma is not (2).

The thesis about the differentiated labeling of interrogative pronouns ma "what" and mi "who" under the influence of the animateness factor of the referent, as well as the influence of this factor on the frequency of labeling of the nominal group in the position of the direct object as a whole, was convincingly proved by P. Beckins on the material of the Hebrew language ^[8], however, as far as the author knows, it has not been previously demonstrated based on the material of modern Hebrew.

Based on the material of modern Hebrew, this thesis finds no refutation, both in grammatical descriptions (which postulate the differentiation of the object labeling of these pronouns, but do not indicate the parameter "animateness" as a factor of variation ^{[12, pp. 270-279]}), and according to the corpus HOT: the interrogative pronoun mi, encoding the animate referent, is marked obligatorily, and the interrogative pronoun ma, encoding an inanimate referent, is marked in isolated cases (1/62). Moreover, the only example of deviation from the general principle recorded by the corpus is most likely the result of the action of additional discursive-pragmatic factors, since it is used in a dialogue where the pronoun ma, pronounced by the addressee in the question, is repeated by the addressee as an "echo" in the answer (see the phenomenon of "topicality"). Nevertheless, the animateness factor, according to the results of the corpus analysis, is undoubtedly the main factor regulating the asymmetric object marking of interrogative pronouns in Hebrew.

Relative pronouns

Pronouns mi/ma can also act as relative pronouns that serve as the vertices of determinative subordinate clauses and encode the patient participant of the situation. The distribution of relative pronouns by the method of object labeling is presented in Table 2.

Table 2.

Thus, despite the undoubted relevance of the animateness factor for interrogative pronouns mi/ma, in the case of relative pronouns, the situation is more complicated, in particular, object labeling of the pronoun mi encoding the animate referent is not mandatory (67/79, 85%). Moreover, a significant variability (33/50. 66%) is also observed in the case of marking inanimate referents encoded by the pronoun ma. Accordingly, we can state that in addition to the animateness factor, which explains the high probability of object labeling of the relative pronoun encoding the animate referent, there is an additional factor (or several factors) regulating the encoding of the referent in such cases.

This factor cannot be the grammatical certainty of the referent, since there are no indicators of determination for relative pronouns.

Apparently, the restriction on labeling such referents is due to a more subtle mechanism that reflects the speaker's opinion about the reality/unreality of the referent within the relevant denotative space, i.e. with a referential status corresponding to a specific pronoun in each specific context. By reference in this study, we mean "the correlation and correlation of linguistic expressions with non-linguistic objects and situations" ^{[4, p. 79]}, or in other words, with reality, or rather with the model of reality in the minds of participants in the speech act ^[5].

Let us clarify that in order to label relative pronouns encoding a patient participant in a situation reflected in a transitive clause, in accordance with the theory of reference, it is necessary to observe the presumption of the existence and uniqueness of the specified referent. Even if these characteristics are not an absolute fact, but are modeled by the speaker's representation of reality, as indicated in the definition above. Let's compare two examples:

In example (3), we see the deictic use of the pronoun ma, since it clearly indicates a specific and unambiguously identifiable referent known from the situational context and, possibly, from the preceding passage of the discourse. Example (4) shows the opinion of a resident of the Bar Yehuda district in Kiryat Malachi about members of his community who come from Ethiopia. The referent of the construction "what they do there" in the general case can hardly be recognized as concrete, uniquely identifiable by all participants in a communicative situation, especially if it is mentioned in the discourse for the first time, which is the case in this case. But the speaker in this example is convinced that there is a certain set of actions of all residents of Ethiopia (which causes him indignation and other negative emotions, which is obvious from the context of the statement), which, in his opinion, is sufficiently specific and obvious for all participants in communication to mark in the position of a direct object.

An important role in the speaker's choice of the method of encoding a direct object, in addition to the correlation of a particular referent with reality or a representation of it, is played by a special referential status arising in the so-called "distributive" context, or, according to A. D. Shmelev, in a "variable denotative space", which "is set by alternately selected elements of some set introduced into consideration", for example, "A horse sometimes knocks down a rider" ^{[6, pp. 98-99]}. Based on the data at our disposal, if the context of the utterance allows us to apply the presumption of existence and uniqueness to the referent of a relative pronoun encoding a patient participant in the situation, then even in a variable denotative space such a referent will also be encoded as a concrete reference object, i.e. marked in the position of a direct object (5).

In example (5), it is impossible to uniquely identify the referent of the description "the one who brought them to Israel" and correlate it with a specific person, including because, as follows from the context, different workers (probably in groups) were "brought" by different employers. However, in this statement, a model of reality is built, where for each worker "brought" to Israel there is a single employer who provided him with entry to Israel, thereby providing a presumption of existence and uniqueness. Thus, within the framework of this denotative space, the attributive description mi she-hevi ' otam la-' arets will be marked based on referential characteristics.

A special case of the referential use of relative pronouns is also a pseudo-dialogical context in which the addressee addresses the audience as a member of the communicative act who is directly present. In these cases, unreal situations in which there are various referents related to the audience (or to the generalized group "speaker + audience", realized through the use of pronouns "we", "our", etc.), are interpreted as real, i.e. they form a model of reality in which the addressee addresses directly to the real interlocutor., and the objects mentioned in this context acquire a reference status and, accordingly, are marked. For example, a commercial website publishes recommendations on what exactly should be purchased as a gift to the mother (6):

Despite the fact that the text of example (6) is addressed to an infinite set of potential buyers, each of whom has a referent of the description 'et ma she-be-' emet rtsitem 'what you really wanted' will be different, the context implies the uniqueness of the referent for each buyer and such distributive certainty allows you to label this referential expression.

An important factor for determining the referential status of relative pronouns is also the speaker's obvious or hidden reflection of his confidence/uncertainty in the reality of the designated referent or the expression of doubt about this fact. For example, the use of restrictive constructions ("at least in our humble opinion") as well as other phrases expressing the addressee's doubt ("I don't know if this is true") or denying the reality of the above ("although I don't think so") correlates quite regularly with the non-referential use of relative pronouns. Based on the trend noted to us, the referential status of relative pronouns is determined, among other things, based on the degree of confidence of the speaker in the reality of the referent in accordance with the following classification: reference status is assigned if the referent is defined as real for the speaker, non—referential - when the speaker identifies it as unreal or expresses doubt about its reality.

Confirmation of the effectiveness of this approach was the high level of significance of the correlation between the reference status of relative pronouns and their marking in the position of a direct object, calculated on the basis of data from Table 3 (p-value < .00001).

Table 3.

Thus, based on the high level of significance of the revealed correlation, the referential status can be considered one of the factors licensing the asymmetric object labeling of relative pronouns encoding in modern Hebrew the O-participant in the transitive clause. However, within the framework of a multifactorial approach to differentiated object labeling, the combination of two factors-licensors, namely the referential status and animateness of the referent, allows us to obtain even more unambiguous results, reflected in Table 4 (p-value < .00001).

Table 4.

The data obtained clearly demonstrate that the referential descriptions encoding the animate referent are marked in 100% of the contexts fixed by the corpus (67/67), encoding the inanimate referent — in 85% (33/39), while non-referential, both animate and inanimate, demonstrate a ban on labeling in the position of a direct object (0/12, 0/11, respectively).

Conclusions and conclusion.

In the study, using the methods of quantitative and comparative analysis based on the material of the experimental corpus of modern Hebrew, on the example of interrogative and relative pronouns encoding the O-participant in the transitive clause, a correlation was revealed between the variability of asymmetric object labeling in modern Hebrew and two licensing factors-licensors: the referential status of the object and the animateness of the referent encoded by it. It was concluded that the variability of object asymmetric labeling of referents encoded by interrogative pronouns, in the vast majority of cases (98%) is regulated by the animateness of the referent, while the variability of labeling of referents encoded by relative pronouns is licensed by both animateness and the referential status of the object. It is the influence of the second factor-the licensor, the referential status, that explains the difference described above in the labeling of interrogative and relative pronouns encoding the patient participant of the situation: asymmetric object labeling of interrogative pronouns is licensed, in the vast majority of cases, only by the animateness parameter of the referent (100% labeling for animate, and 2% for inanimate), whereas object labeling relative pronouns are licensed by a combination of the parameters "referential status of the IG" and "animateness of the referent".