SUBJECT-TRACKING AND TOPIC CONTINUITY IN THE CHURCH SLAVONIC TRANSLATION OF THE STORY OF ABRAHAM AND HIS NIECE MARY Текст научной статьи по специальности «Языкознание и литературоведение»

Subject-tracking and Topic Continuity in the Church Slavonic Translation of the Story of Abraham and His Niece Mary

Установление субъекта и непрерывность темы в церковнославянском переводе Слова об Аврамии и его племяннице Марии

Dieter Stern

Ghent University Ghent, Belgium

Дитер Штерн

Гентский университет Гент, Бельгия

Abstract

The present article addresses issues of referentiality and text cohesion in a Church Slavonic narrative text. Starting with the specific problem of referential conflict as formulated by Kibrik [1987], issues of tracking personal participants in a narrative text are broadly explored in order to arrive at a rationale for the construction of cohesive text interpretation through topic continuity in subject position. The article takes an interpretative text-based approach of close-reading and argues for participant tracking to be dependent on text genre and general cultural prerequisites of text reading and interpretation rather than on systemic grammatical features of language. It is also hinted at the possibility that medieval narrative text genres (like the Byzantine-Slavic

Цитирование: Штерн Д. Установление субъекта и непрерывность темы в церковнославянском переводе Слова об Аврамии и его племяннице Марии // Slovene. 2019. Vol. 8, № 2. C. 57-86. Citation: Stern D. (2019) Subject-tracking and Topic Continuity in the Church Slavonic Translation of the

Story of Abraham and His Niece Mary. Slovene, Vol. 8, № 2, p. 57-86. DOI: 10.31168/2305-6754.2019.8.2.2

This is an open access article distributed under the Creative Commons Attribution-NoDerivatives 4.0 International

hagiographic genre being explored in this paper through the specimen of the Story of Abraham and Mary) may adhere to a type of narrative construction which places more responsibility on the reader-listener than on the narrator.

Keywords

topic continuity, Church Slavonic, hagiography, Bdinski sbornik

Резюме

В статье исследуются вопросы референциальности и непрерывности темы в одном церковнославянском тексте — в Слове об Аврамии и его племяннице Марии. Исходя из особой проблематики референциального конфликта, разработанной Кибриком [1987], вопросы идентификации и установления личных участников повествования трактуются более широко, благодаря чему оказывается возможным моделировать общие предпосылки интер-претативного построения непрерывности темы. На основе целостного подхода к отдельному тексту автор приходит к выводу, что установление участников повествования зависит не от системного устроения языка, но от жанровых свойств текста и общекультурных знаний слушателя-читателя. Предполагается, что средневековые повествовательные жанры (такие, в частности, как славяновизантийская агиография, анализируемая в настоящей статье) принадлежат к такому повествовательному типу, который перекладывает ответственность на расшифровку текста с повествователя на слушателя-читателя.

Ключевые слова

непрерывность темы, церковнославянский язык, агиография, Бдинский сборник

1. Introduction

There is probably no narrative text, which does not have to address the basic technical problem of keeping the figures and items, or technically spoken the topics it introduces apart throughout the narrative. This problem of language processing and text interpretation has been addressed as referential ambiguity or conflict [Kibrik 1987; 2011: 62-67], and it is this type of interpretive insecurity, which inspired this article. On reading the story of Abraham of Qidun and his niece Mary (AQM), which provides the basis for our analysis, a modern reader will find keeping track of the story's figures, though there are not a lot of them, time and again a daunting task, though in the end it turns out that at no point throughout the story ambiguity really prevails.

For the readers of this article, who will, as a rule, not be acquainted with the story of Abraham and Mary, a short summary of its plot will be useful in order to be able to follow the argument. Late in life Abraham, a 4th century hermit in the Syrian desert, took charge of the daughter of his recently deceased brother, a girl of 7 years by the name of Mary. Abraham makes up his

mind to raise her as a hermit and teach her the essentials of a hermit's life. Mary followed Abraham's example in all respects, but by the time she came of age, her outward beauty attracted the attention of a young man pretending to be a hermit too. This scoundrel succeeded in seducing Mary to commit with him an act of 'dirty impiety'. After having committed her sin, Mary repents and leaves her cell out of shame, unbeknownst to her uncle. In a long monologue she bewails her fate and argues that now that she failed so utterly she could never expect to find forgiveness and grace anymore, neither from her uncle nor from God. As if to stress her utmost degradation she enters a brothel to become a whore. Only two days after Mary left him, Abraham came to realise through a dream that his niece is no longer with him. He then asks the help of an acquaintance to go and find out her whereabouts. After two years her secret abode is discovered and made known to Abraham, who in the guise of a soldier leaves his cell to go and bring Mary back. In the seclusion of a private chamber within the brothel, Abraham makes his true identity known to his niece, who bursts into tears and will not believe that there still is hope for her. Finally, she returns with her uncle to the hermitage. Both lead henceforward a life in strictest austerity.

2. What's in a topic?

For any figure or item being introduced into a text to become a topic, specificity and individual identifiability are required. Individuation in narrative context is prototypically linked to properties like animacy as a prerequisite for becoming a protagonist, i.e. becoming a binding and focal element determining the cohesion of the narrative. Accordingly, abstract concepts are less prone to individuation than animate referents (ideally namable individuals acquiring personhood). Therefore, inanimate abstract concepts, inanimate concrete entities, animate entities, and namable persons show accordingly increasing likelihood of becoming promoted to the status of continuous topic within narrative discourse. The highest probability for acquiring topicality is found with fully individuated and namable persons, which are treated by the anthropocentric human cognition as more prototypical compared to other types of referents [Kibrik 2011: 4, 406]. It should also be noted that topic, subject, and agentive case role are prototypical-ly correlated [Givon 1983: 20-23]. This correlation is intrinsically motivated: agents keep the story going; in order to do so they will have to appear continuously throughout the story, which will make them topical. Both properties are causally interlinked and find their common formal expression as subjects. This means that subjects typically code acting topics, as opposed to non-acting topics, which are usually being coded in other syntactic roles.

As for namable persons, an additional line should be drawn between actors on the scene and persons not directly or only marginally involved in the evolu-

tion of the plot being reported.1 The first rank certainly highest in topicality by dint of their capability of acquiring true protagonisthood [Kibrik 2011: 406]. As 'thematic' participants [Levinsohn 1978: 75], i.e. as participants around whom a paragraph or episode is organized, they will usually initiate episodic action and perform mostly in subject position throughout the episode. We assume protagonisthood to operate at different narrative levels (text vs. episode within the text), which may interact in determining the prominence of a participant. Especially in hagiographic texts, like AQM, there is a very pronounced difference between figures being identified in the title or rubric as the saint for whose sake the text was written, and all remaining figures. The first may be supposed to be present throughout the whole narrative and dominate most episodes, while the latter are usually restricted to a limited number of episodic sequences within the narrative. The saint functioning as the nominal hero of the whole narrative is not only a ubiquitous actor on the scene, he or she is also the one through whose eyes the reader-listener will enter the story. This singles him or her out as the one person being closest to the audience, which means that he or she more than any other participant will be on the audience's mind and therefore will need explicit referencing less than anyone else. Within AQM the role of most prominent protagonist is clearly assignable to the titular saint of the story, i.e. Abraham, but his niece Mary acquires a role almost equal to his as a co-protagonist. Both protagonists are in fact the only figures which will continue their presence across episode boundaries, whereas all remaining figures are restricted to just one episode. Abraham's higher rank in protagonisthood in relation to Mary may possibly be gleaned from the fact that he is the only participant who repeatedly enters new episodes by way of unmarked zero2 referencing (see below, section 3.2, on episode boundaries).

Though topicality may be least likely for inanimate (abstract and concrete) referents, it cannot be excluded for them on principled grounds. But it appears that within the narrative of AQM all inanimate nouns (abstract as well as concrete) appear to be non-topical, or more correctly only punctually topical in the sense that they are never anaphorically referred to throughout the narrative and thus never form continuing topics. Could it be that within the narrative hagiog-raphic genre to which AQM belongs, zero anaphora is reserved for persons, or

1 Major vs. minor participants according to Levinsohn [1978: 69] or participants vs. props according to Grimes [1975: 43-45].

2 In the discussion that follows the least marked option for subject coding in Old or Church Slavonic (henceforward just CSL without differentiation between the two) will be persistently addressed as zero anaphora and the like. Speaking of zeros is, however, in our case not exactly true. As so many other European languages, CSL codes obligatorily referential information about subjects on the verb through personal desinences. True zero anaphora would imply that no information is provided at all, like e.g. in Japanese or Chinese. Other terms have been suggested, such as 'bound pronoun' as opposed to 'free pronoun' [Kibrik 2011], or gramm-index [Haspelmath 2013: 206-207].

even more narrowly for actors on the scene? As a matter of fact all cases of zero (subject) anaphora in AQM refer to acting persons, which means that only these acquire topical status within the narrative. The conclusion that referencing in AQM is not sensitive to inanimate subjects may also be gleaned from the fact that intervening inanimate subjects, i.e. subjects that visibly interrupt reference chains, never effect the choice of more explicit means of referencing. Continuing acting topics are resumed regularly by zero after intervening inanimate subjects, the latter being ignored or skipped as possible referencing sites.

Not all persons that make their appearance in AQM acquire true protago-nisthood. Of the 19 persons or groups of persons that are mentioned throughout AQM only 6 temporarily take on the role of interactants on the scene of reported events. Accordingly, they become subject to referencing by zero - or occasionally also pronominal - anaphora. All of these second-degree protagonists, except the devil, take their part in the unfolding of the plot on the immediate scene of action. The devil, but interestingly enough not God, becomes topical for being assumed to play an active part behind the scene and governing the sequence of events. In a way he thus appears more real than God by way of personal commitment to immediate action in this world. Ultimately, a basic requirement for topicality within the narrative hagiographic genre is direct involvement in the plot, and the more central a figure's position in the plot is, the more likely he or she will be referred to by zero anaphora.3 This is not to mean that protagonisthood determines zero reference in a way that it is predicted by and reserved to this status. It is rather that high percentages of zero referencing are indicative of protagonisthood that is brought about by a dominant presence on the scene, which in its turn causes a high activation status throughout the narrative. We also assume that in cases of referential conflicts being solved by identifying one of the competing participants by means of nominal resumption, it will be the less topical referent that will undergo nominalization.4

3. Tracking subjects in AQM5 3.1. Subject tracking devices: zeros, pronouns and nouns A fully-fledged functional linguistic device for participant tracking would require that formal means for keeping topics apart are made available throughout, irrespective of the potential for ambiguity. For such a system iconicity may be expected to hold, as specified by Givon [1983: 18] for topic marking.

3 This confirms the general findings made by Marslen-Wilson, Levy & Tyler [1982].

4 This assumption is strongly suggested by the material for referential conflict presented and discussed by Kibrik [1987: 132].

5 The analysis that follows is in part indebted to the insights gained from extensive discussions with Joachim Gaelens during a course on a pertinent topic at the University of Ghent in the academic year 2016-17. In the meantime, Joachim Gaelens completed

The more accessible a topic is to the reader-listener, the less phonetically marked will be its formal trace and vice versa. The least marked form imaginable, viz. zero anaphora is accordingly iconically associated with the best accessible topic imaginable, viz. an immediately continuing topic with no potentially competing topic around. Beyond this general principle of iconicity, each language will show its own preferences to apply the variety of referencing alternatives available by either over- or undercoding continuing topics. It appears that modern colloquial English storytelling appears to operate on strict rule-like assumptions about the scope of referential means. Thus, pronouns and zeros are regularly interpreted to generally refer back to the closest nominal antecedent, which results in a fairly predictible and regular use of lexically specific anaphoric elements [Marslen-Wilson et al. 1982: 352-354]. In stark contrast to this, CSL - at least as represented in AQM - seems to tend very clearly to the undercoding end with no particular preferences for the reference of unmarked anaphors. CSL appears to operate on the basis of a low-cost default-system, which will activate marked forms only very sparingly. The default for given topic resumption is zero anaphora for all cases of same-subject (with or without intervening non-topical different subject), but also for different-subject situations, if narrative and commonsense knowledge and/or indexical cues6 are sufficient to identify the participant. 7 Nominal resumption (NR) is used for the introduction of topics-to-be and reactivation of remote topics. Remoteness seems to be treated in terms of episodic prominence of a protagonist rather than a measurable distance in words. Thus, the speaker of a lengthy stretch of monologue will be resumed, as he ends his monologue and remains the acting person of the immediately following sentence of the narrative proper, by zero anaphora irrespective of the number of the intervening

an MA thesis on the topic, in which he took a somewhat different stance and approach [2017]. Both the MA thesis and the present paper reflect our joint work on and discussion of the AQM-text. Without this inspiring collaboration the article in its present form would not have been possible, for which I wish to express my gratefulness to Joachim Gaelens. My gratitude also goes to the participants of the 12th Meeting of Paleoslavists at Vienna on September 29-30, 2017, especially to Florian Wandl, who commented on the presentation of my findings and provided additional hints for the final elaboration of this paper.

6 By indexical cues we mean formal cues which allow for the identification of referents, but which do not serve primarily participant tracking. Kibrik [2011, ch. 8 & 9] treats these indexical cues as 'referential aids'. A typical feature of this type of cues

is that their availability cannot be counted on in every situation which would require disambiguation. For example, morphology specifying gender helps a lot where concurring prominent topics happen to be oppposed by different gender, but in the case of two competing referents having the same gender the reader-listener will be left all by him- or herself.

7 In this respect, CSL resembles Chinese, for which Li & Thompson [1979: 322] observe that non-occurrence of anaphoric arguments is the unmarked case, which may be expected anywhere anytime. It is full resumptive forms which require explanation.

sentences/utterances, as long as both monologue and subsequent action form part of the same episode within the narrative. It is as if intervening passages of direct speech were skipped, as it were, in participant tracking on the level of the narrative. Even with longer intervening stretches of direct speech like Mary's lament after having been seduced, zero anaphora will be employed, if the subject of the sentence resuming the narrative is coreferential with the author of the utterance. As the subject of a sentence which immediately precedes direct speech is, as a rule, the speaker of the utterance itself, it is quite obvious that, though the narrative flow itself is being interrupted, the referent remains in the focus of the reader-listener's attention as the person delivering the speech. More often than not, the speech event itself will contain direct and indirect hints as to the identity of the speaker, thus reinforcing his or her status as the most prominent topic entity with the reader-listener. Once again it becomes evident that it is not so much a discrete formal subject unity which is being referred to.

Pronominal resumption (PR) in subject position is restricted to a very small set of narrowly delimited cases. There being no 3rd person subject pronouns proper in CSL, the only resource to resort to for anaphoric referencing consists of the deictic pronouns сь, тъ and онъ.8 Throughout AQM there is only one instance of тъ and two cases of сь being used als resumptives in subject position:

(1) Разоумнымь ибо и Дховнымь си дивна соуть, испльн же ползе и покааша [BS

fol. 1r].

For the wise and spiritual these things are wonderful, full of benefits and penitence.9

8 Vaillant [1964: I, 141, §93] identifies mz as filling in the paradigmatic gap in the nominative for anaphoric u as an emphatic form. Contrary to this, Leskien [91969: 98, § 78] makes onz fill in the same gap in his paradigm of anaphoric kio, whithout further comment. Lunt [2001, 63, § 4.25] combines both claims by stating that the "function as third person pronoun [in the nominative] is taken by a demonstrative, usually tz 'that one', less often onz 'that one yonder'." Bielfeldt [1961: 148] is the only one to assert a true paradigmatic gap for the nominative of the anaphoric pronoun. Judging from our observations on AQM neither mz, nor onz can in any meaningful way be described as a regular suppletive within the paradigm of anaphoric u, kio. See also Kibrik [2011: 260] who clearly states that the only reduced referential devices for subject position are verbal desinences. Vecerka, Keller & Weiher [1993: 51-59] identify and discuss several instances of mz, ct and onz used as "semantically neutral" anaphoric subject pronouns. They conclude that in all cases the three demonstratives preserve part of their deictic function, so that they cannot be considered part of the 3rd person pronoun paradigm.

9 All translations are taken from Sebastian Brock's rendering of the Syriac original [Brock & Harvey 1987: 29-36], which by and large is remarkably close to the Slavic text, so that only few adaptations had to be made to make his translation fit the Slavic text. The Slavic text is available in print as a facsimile edition [Dujcev 1972: 1r-17v] and a regular print edition [Scharpe, Vyncke 1973: 43-55]. For a digital edition of the Bdinski sbornik version of AQM see http://bdinski.obdurodon.org/.

(2) шнь же разоум'Ьвь ако то еи есть быти ... [BS fol. 6r]

He, being convinced that this was she...

(3) се бо честь кс неприазниннаа. [BS fol. 13r]

This belongs to the Evil One.

In none of the three cases is there a formal antecedent, with which the deictic would agree in gender and number. All three forms are neuter. From this it can already be gleaned that the reference is generic rather than specific. The neuter plural deictic си in (1) happens in fact to be cataphoric rather than anaphoric, referring to the whole of the story which is going to be told. The neuter singular то in (2) refers to the description of Mary given to Abraham by his acquaintance. Neither reference is topical in any possible sense of the word. Unlike these two cases, the neuter singular се in (3) indeed refers back to the fairly discrete unit мало злата и ризы ('a little gold and clothing') introduced a few sentences earlier. Thus, (3) is the only case which qualifies as topical in the broad sense of identifying a unit that is at least in theory capable of becoming a topic entity within the narrative. It ought to be noted that this is also the only case throughout the narrative, where a non-animate unit is referred back to.10 For the rest, the deictics сь and тъ are used only attributively within nominal expressions such as блаженный сь or блаженная та, in which contexts they deictically support resumption, but are not themselves primarily resumptive.

As for the deictic онъ, independent use in subject position is restricted to two areas of application: (a) to mark off the respondent to a summons in adjacency pairs,11 and (b) in contrasting comparisons, where онъ is regularly opposed to correlative самъ. Contrasting comparisons represent a very small and specific subset of the wider array of narrative situations requiring disambiguation of some kind in order to keep track of the main participants. Though contrasting comparisons testify to the utilisation of pronominals for obviation, this can at best be called embryonic.

In adjacency pairs there is, as a matter of fact, no primary disambiguation problem at hand. Principally, once the identity of the interlocutors has been determined nominally in the starting sequence of a conversation, the rigid pattern of the recurrent schema of summons and response organized in adjacency pairs will structurally define the identity of the interlocutors throughout the rest of the conversation. Only with longer sequences of dialogic exchanges the reader-listener would run the risk of losing track of who is saying what,

10 This is in line with the observation of Vecerka, Keller & Weiher [1993: 58] that ct seems to be the only pronoun capable of representing inanimate subjects.

11 This specific usage has also been observed for other CSL texts by Vecerka, Keller & Weiher [1993: 55].

especially if the conversation does not consist of easily identifiable speech act categories like e.g. questions and answers, which neatly reproduce the underlying summons-response schema on the visible/hearable surface. In conversations being embedded in the narrative, the deictic онъ primarily safeguards the visualization of this basic underlying structure of the dialogue, thus helping the reader-listener to keep track of the dialogue structure and by the same token of the identity of who is saying what.

In AQM dialogues are embedded into the narrative by means of verba di-cendi applied on the narrative level to mark off the beginning of direct speech. As the embedded dialogue starts, the sentence heading the dialogue introduces both interlocutors either nominally or anaphorically depending on their topical status within the episode of which the dialogue forms part. In (4) Abraham (ABR) is introduced as the summoner by means of a zero anaphora, because he happens to be a prominent topic already.12 There is even same subject reference in relation to the immediately preceding sentence. The inn-keeper (hos), however makes his first appearance on the scene and is accordingly introduced nominally as a respondent.

(4) вьсклабив же се рече 0 (ABR) кь гостинникоу (hos): [BS 7r] He spoke to the tavern keeper with a smile:

In (5) the summoner MAR is introduced nominally as дтвица, because this sentence marks also the beginning of a new episode after a long intervening commentary, thus requiring the protagonists to be reactivated by nominal referencing. Note also that the respondent is in this case not indicated in the opening sentence as indirect object of the verbum dicendi, but instead by an admittedly unspecific vocative phrase in the speech event itself:

(5) по вьзвеселкныи же их реч двца (MAR):

"господине (abr), вьниди на ложе да поспив'Ь тоу." [BS 10r]

After they had chatted together, the girl said: "Please come into my bedroom so that

we can sleep together."

12 The persons or groups of persons appearing throughout AQM are: Abraham (ABR), his niece Mary (MAR), Abraham's brother and Mary's father (BRA), Mary's other relatives (REL), the monk who seduces Mary (MON), Abraham's relative who is sent to find out about the whereabouts of Mary (REC), the inn-keeper (HOS), all of them being immediately involved in the ongoing plot, Abraham's friend Efraim (EFR), who is supposed to exist on the level of the reality of the plot, but who aquires no acting part in it because of his physical absence from all scenes of action, God (GOD) and the Devil (DIA) who perform as metaphysical actors behind the scene, the symbolic figures of the evil serpent (SER), the pure dove (COL) and the ravenous wolf (LUP), who make their appearance in Abraham's visions as well as in his moral judgements, and finally idealized or literary figures invoked for purposes of illustration, comparison and judgement, viz. the ideal shepherd (PAS), the ideal scout (SPY), the biblical namesake of Abraham (ABR1), Lot (LOT) and the kings (REG) from whose grip Lot was liberated.

Once both interlocutors have been identified, a rigid pattern for the subsequent adjacency pairs is being observed, as long as the conversational flow is not interrupted by additional narrative elements or comments, which may cause nominal reactivation of one or both interlocutors. In cases where no specific conditions prevail, the summoner is referred to by zero anaphora and the respondent by онъ by default, as in the following example of a second turn-taking in a conversation between ABR and MAR:

(6) Гла 0 (ABR) ки (mar):

"На мн'Ь гр'Ьхъ твои, чедо мок. {...}" Wna (MAR) же реч кь нкмоу (abr): "Аще в^си, ако могоу се покаати {...}" [BS 12v]

He said to her: "My daughter, I have upon myself your wrongdoing. {...}" She said to him: "If you know (for sure) that I can repent, {...}."

The default assignment of zero anaphora for the summoner and of онъ for the respondent operates on the basis of metaphorical deixis. Zero anaphora as unmarked form can be taken to be inherently maximally proximate, whereas distal онъ 'that one yonder' marks the extreme opposite in terms of deixis. It is, as if the reader-listener looks at the dialogue (and possibly also at the surrounding scenery) from the perspective of the summoner, who by initiating the dialogue becomes the focal center of it.13 Accordingly, being the starting point of the adjacency pair, the summoner himself is visualized as being close to the reader-listener, while the respondent is located at some metaphoric distance from him. It is the same metaphoric use of deixis, which is at the bottom of obviation. In languages having obviatives these will usually be opposed to so-called proximate forms [Payne 1997: 212-213; Kibrik 2011: 311-312]. Proximates represent participants which are more central to the story and therefore closer to the reader-listener, whereas obviatives will mark participants of minor importance to the story and hence to the reader-listener as well. As Marianne Mithun [1996: 146] puts it, "the proximate participant is the one from whose point of view events are presented." Her description neatly fits the distinction of proximate zero anaphora as opposed to obviative онъ observed for the sequencing of dialogue exchanges in AQM. Thus, there are indeed formal means of obviation in CSL, but their use is severely restricted. AQM offers only one example in which orn appears to be used obviatively outside of the direct speech sequencing domain:

13 The assumption that the reader-listener is as the experiencer of the story spatio-temporally involved in the story by taking on the role and perspective of the story's acting figures, especially that of the main protagonist, is supported by findings within the field of cognitive linguistics (see e.g. Zwaan [2003: 43-44] on 'spatial region' and 'perspective', and 55 for further references).

(7a) Шедше тоу абик посланны (REC) оув'Ьд'Ь изв'Ьсто w нки (mar), вид'Ьв 0 (REC) же ю самоу (mar).

(7b) Пришедь 0 (REC) же повода 0 (REC) емоу (abr) сице коегождо знамений начрьтаник.

(7c) Шнь (ABR) же разоум'Ьвь йко то еи (mar) есть быти, реч 0 (ABR) же пршести кмоу (abr) воиньскоую подобоу, кон же и wроужик, и wтврьзь 0 (ABR) дверце хизини, и изл'Ьзе 0 (ABR), wблькъ 0 (ABR) се вь воиньскыи wбразь. [BS 6r]

Having gone, the man, who was being sent, found out about her and saw her, whereupon he returned and gave him (Abraham) an outline of each sign (of her whereabouts). Convinced that this was she, he asked to have brought to him an outfit of soldier's clothing, a horse and weaponry, and then he opened the door of his home and went out, having put on the military dress.

But this again is a dialogue sequence, clearly recognizable from the verba di-cendi (first turn: повтда > second turn: &нь же ... ре11), the only difference being its making use of reported rather than direct speech.

It looks like CSL has not worked out any generally applicable device to cope with referential ambiguity outside sentence limits beyond possibly the fundamental opposition of nominal resumption (NR) vs. zero anaphora. Within the admittedly narrow confines of AQM, none of the deictics discussed can be considered as freely available anaphoric elements for subject position being employed to build up meaningful functional oppositions with zero anaphora on the one hand and NR on the other. Besides, CSL seems to treat the risk of subject ambiguity very negligently into the bargain.14 NR appears not to be used in a regular and systematic way to effectively prevent ambiguity, let alone to minimize its risk. The basic formal contrast between nominal and zero subject resumption reflects a rough differentiation between proximate or prominent topics, which would not need explicit reference, and distal and remote topics, that can be expected to be reactivated by NR. The prototypical environment for zero anaphora would accordingly consist of same subject environments, whereas NR ought to prevail in different subject environments. This is reflected to some degree in the numerical distribution of zero vs. nominal anaphora throughout the text of AQM:

14 On comparing CSL to the nine case studies presented in Givon [1983] CSL turns out to represent a rather extreme case. Japanese is the only language which deals with topic discontinuity in a similarly restrictive way as CSL. Both languages employ zero anaphora as a default device, restricting the use of stressed independent subject pronouns to the very limited subset of contrastive expressions. It must, however, also be emphasized that CSL unlike Japanese has obligatory person marking for subjects on verbs, that in the end makes it much less elliptic than Japanese, which really seems to be a very extreme case [Hinds 1983: 49-50].

Reference zero anaphora pronominal resumption (PR) nominal resumption (NR)

Same Subject 168 (96%) 4 (2.3%) 3 (1.7%)

Different Subject 42 (47%) 16 (17%) 32 (36%)

All subject contexts15 210 (79%) 20 (8%) 35 (13%)

Whereas same subject environments very clearly confirm the proximate interpretation of zero anaphora, the picture is less clear for different-subject environments. PR being restricted to a few rigidly defined contexts of application, one would at least expect a marked preponderance of NR with different subject environments. The actual numerical dominance of zero anaphoras over NR would run counter to expectations derived from a proximate-distal interpretation of the contrast of both formal devices. As a matter of fact, eight occurrences of zero anaphora may indeed be subtracted from the list. For these cases it can be shown that the immediately preceding subject is nontopical and may be discounted as a potential point of reference, so that ultimately accessibility may be assumed to be no lower than in a same subject environment proper. This is illustrated by the following sequence of subject switches from the prominent topic entity MON - i.e. the monk who is about to seduce Mary - to nontopical COR back again to MON. It should also be noted that MON is present in (8b) as well, so that the chain of reference remains even formally intact:

(8a) Видоваше 0 (MON) же и блаженоую сию дверцами, иже (MON) бывь неистовь хот'Ьше 0 ки бесЬдовати.

(8b) акоже бо пламень раждеже се срдце (COR) кго (mon) неистовнык любве.

(8c) Желаа 0 (MON) же ек доволно вр'Ьме ако л'Ьто едино, дондеже раслабыть свои помысль. [BS 3r-v]

He also happened to see the blessed girl time and again through the door, and fell in love with her and wanted to keep company with her. His heart burned like aflame because of his mad desire (for her). He longed for her quite some time, until he succeeded in softening her firm resolve.

15 The marginality of pronouns for 3rd person subject marking in CSL (AQM) can be gleaned from a comparison with the distribution of the same categories in narrative discourse in Modern Greek, a prodrop language which at least has a specific, i.e. non-deictic 3rd person subject pronoun. The scores for ellipsis range from 37.8% to maximally 59.1% in the stories investigated by Georgakopoulou. Pronouns are found in 19.5-26.5 % of all occurrences [Georgakopoulou 1997: 111]. According to Olson [1977] and Chafe [1982] low rates of explicitness, such as found in CSL (AQM) would correlate with oral rather than written text production. It remains, however, to be shown whether their claims about explicitness in relation to mode of production are universally applicable across time and cultures.

The resulting ratio of 34 null subjects against 32 cases of NR, however, does still not satisfy expectations of a clear-cut functional division. It goes without saying that NR constitutes a quite awkward and cumbersome instrument for smooth and efficient participant tracking; due to that avoiding its use and reducing it to the least possible quantity would appear quite natural [Kibrik 2011: 51; Gordon et al. 1993]. What appears to emerge in the end is rather suggestive of a non-system with zero anaphora as the default for all situations, and NR being used only sparingly to meet the most basic requirements of supporting the reader-listener in his efforts to create text cohesion. The following discussion should accordingly focus on how AQM manages to grant an unambiguous reading of the narrative by providing means and cues to track participants for every subject position in the text. Can it be shown that NR will occur primarily where ambiguity of subject reference is seriously at stake? Or is the use of NR determined by other factors? And what risks does the writer-narrator take in order to keep the congestion, which is likely to be caused by a too ample use of NR, as low as possible? Can it be shown that NR is employed to effectively reduce ambiguity and at the same time avoid congestion? And are there possibly other conditions, like episodic structure, text layering, and change of narrator footing, requiring NR? Let us first have a look at a complete episodic unit in order to arrive at a general idea how referent identity is constructed in a typical scene of AQM:

(9a) Б'Ь же н'Ькто име им'Ьк чрьньчскок (MON), иже (MON) хождаше кь блаженомоу йко виною полезною сь тьщаникмь многомь.

(9b) Видоваше 0 (MON) же и блаженоую сию (mar) дверцами, иже бывь неистовь, хот'Ьше 0 (MON) ки (mar) бесЬдовати.

(9c) йкоже бо пламень раждеже се срдце (COR) кго (mon) неистовнык любве.

(9d) Желаа 0 (MON) же ек (mar) доволно вр'Ьме йко л'Ьто едино, дондеже раслабыть свои помысль.

(9e) И wтврьзши (MAR) дверце хызнык и изиде 0 (MAR) к нкм8 (mon).

(9f) И wскврьни 0 (MON) ю (mar) скврьннимъ безаконикмь.

(9g) По сьтворени же гр'Ьховн'Ьмь оужасе се помысль ке (mar).

(9h) И растрьзавши (MAR) рызи свок бийше 0 (MAR) се по лицю и хот'Ьше 0 (MAR) се сама оудавити wx печалы. (3r-v)

Now there was a man who was nominally a monk, who used to come regularly and visit the blessed Abraham with considerable enthusiasm on the pretext of some useful design. He also happened to see the blessed girl time and again through the door, and fell in love with her and wanted to keep company with her. His heart burned like a flame because of his mad desire (for her). He longed for her quite some time, until he succeeded in softening her firm resolve and the girl eventually opened the door of

her cell and came out to him. He bespattered her with the mud of his lust. Once this sinful episode had taken place, stupefaction seized hold of her mind. She tore off her garment, and beat her face and wanted to strangle herself in grief.

The sequence (9a-h) comprises the episode of Mary's seduction by a monk. It is characterised by a high level of interaction between two prominent participants, the monk (MON) and Mary (MAR) requiring a rapid succession of subject switches as the action unfolds. These occur from (9d) to (9e), where MAR takes the leading role from MON, then immediately again from (9e) to (9f), where MON takes over again, and finally again from (9f) to (9h), where the episode is concluded by MAR bewailing her sad fate. Zero anaphora is used throughout for both participants. NR is not employed anywhere in order to keep both protagonists apart. Keeping track of what is going on seems to depend largely on grammatical clues, which one intuitively expects to be fewer in prodrop languages, like CSL, compared to a language requiring pronouns in subject position. The essential difference between pronominal and zero anaphora lies in the abundance of formal indexical cues supplied to find one's way through the maze of the ongoing interaction, which is usually richer with pronouns than with desinences [Grimes 1975: 49]. Whereas zero anaphora will provide no indexical cue other than the information on person and number supplied by the verb desinences, pronominal resumption additionally offers gender information as a cue for identification. Both interactants being singular, gender would be the only feature to keep them apart. Though the prodrop language CSL completely misses out on providing gender information for the subject position, the text is still replete with formal gender cues, viz. on conjunctive participles as well as on nonsubject resumptive pronouns. Thus, the feminine ending of the preterite active participle wmepb3mu in (9e) helps to identify the subject as does the masculine resumptive pronoun k hkm8 in the same sentence, though in a less straightforward manner. Knowing that the person towards whom the action is directed is masculine, i.e. MON, Mon cannot possibly be the agent of the sentence occupying the subject position marked by zero, so it must be MAR.16 Of course, these cues are rather accidental and are not specifically meant to help to identify subjects. But somehow, they can be relied upon to be present more often than not in a way that renders a specifically designed tool for referential subject disambiguation unnecessary in most cases.

Unfortunately, AQM does not provide episodic material where number and gender cues prove insufficient. At the same time, the seduction scene also underscores the rigid cultural restrictions on admissible types of interactive encounters. In a way, cultural restrictions put severe limits on the theoretically

16 This type of inferencing based on logical exclusion has been termed 'the engagement factor' by Kibrik [2011: 294].

possible diversity of interactive configurations. In the end, however, it is not so much the restrictiveness of culture with its diminishing effect on the variety of possible scenarios, which could lead to referential conflict. What is more is the prescriptive nature of culture, which proffers a rich choice of prefabricated and regularly reproduced default patterns for interactive encounters. In the end, succesful participant tracking might be safeguarded at a very different level, indeed: the nominal labels identifying both participants provide the reader-listener with moral judgements, which determine his or her expectations and allow him or her to assign each of the two participants an almost stereotypical dramatic role, that can be associated with prototypical actions. The visiting monk is introduced by the ironic description of 'someone who goes by the name of a monk', thus implying that he is a cheater who pretends to be a monk. On the basis of this description the reader-listener will assign the visiting monk the role of the trickster, an agent of evil and a tool of the devil, whereas Mary being called 'that blessed one' takes on the complementary role of the honest and unsuspecting victim. All of the verbal actions occurring throughout the episode can be assigned in a prototypical manner to either role. Tricksters initiate action (видоваше, хоттше бестдовати, wскврьни), they are easily aroused by immoral desires (бывь неистовь, раждеже се срдце кго, желаа). Victims are made by tricksters to do things they had no prior intention of doing (трьзъши дверце, изиде as a consequence of раслабыть помысль кк), and doing things contrary to their original intention. Only they will have a reason to feel regret in the end about what has been actually happening (оужасе се помысль ке, растрьзавши рызи, биаше се по лицю, хоттше се оудавити). The identity of the co-occurring participants will thus be largely determined by a culturally available schema or script [Schank, Abelson 1977; van Dijk, Kintsch 1983: 47-49; Givon 1992: 15]. It should be noted that for cultural scripts to operate as a participant tracking device it is required that the events reported do not move beyond prototypical patterns of (inter)action. We therefore assume that sudden and unexpected turns will require more specific means of participant identification. The more a particular text or text genre sticks to established and commonly known patterns of (inter)action - and hagiography does so to a remarkable degree - the more will scripts be available to help participant tracking.

This and other episodes teach us that in many instances there is no real need for a device of participant tracking,17 which would formally instantiate a referential conflict filter, suggested as a universal cognitive device by Ki-brik [2011: 63-64], and it may be assumed that CSL does not provide for any

17 At least this is true for AQM. Most episodic events within AQM do, however, not involve more than two interacting parties. Though confusion of referents is possible even with two interactants, the issue of referential conflict becomes obviously the more pressing the more interactants get involved [Grimes 1975: 34].

reliable, i.e. strictly formalized device of anaphoric referencing in cases of referential conflict. It leaves the reader-listener all to him- or herself to make sense of the narration and construct it as a cohesive whole. The reader-listener of the CSL translation of AQM is required to be equipped with a rich cultural knowledge about literary figures and types, and the capability to apply them on the spot by simultaneously decoding the plot and assigning it an interpretation at a deeper, mostly moral level. At the same time, he or she is expected to keep track of all events of the story so as to be able to situate every single event being reported in the larger framework of the unfolding plot.18 Cultural and narrative knowledge, however, presupposes an experienced public that will be able to make sense of a story with as little grammatical or other formal support as possible. The presence of linguistic features and the forms they take, and the ways they are applied would thus appear to be dependent on genre properties, where genre is understood as a specific textual format, that is subject to specific social and cultural conditions. It is these conditions, which act as patterns and horizons of expectations in restricting referential choices.

Reference, as it is seen at work in the seduction episode, does not operate on the level of grammar, not even of language. What we called zero anaphora up to this point may turn out to be no anaphora at all, but just absence of formal reference where it is not needed, because cohesion is constructed by fitting the narrated events into a ready-made framework of cultural and situational expectations as well as common knowledge. Thus, there would be no reference chain, which links all the zero elements to their antecedents.19 What we see instead, is a process of sorting and classifying events according to culturally determined role patterns or scripts, and assigning them to the acting protagonists, accordingly. Cohesion becomes here more than anywhere else apparent as a construct or product of the interpreting reader-listener, who takes it for granted that texts are coherent much in the same way as in conversation the interlocutors will assume that every contribution is relevant with respect to the purpose or direction of the conversation.20 Cohesion in the passage above is literally created out of nothing, linguistically speaking.

Episodes like the one analysed above do not form an exception within AQM: there are others very much like it, but usually grammatical indices and knowledge of the narrative situation are sufficient to deal with them. Only

18 For a similar analysis of a specimen of narrative discourse which arrives at basically the same conclusions, see Marslen-Wilson, Levy & Tyler [1982: 362-364].

19 For a similar assessment of zero anaphora see Marslen-Wilson, Levy & Tyler [1982: 365], who argue that zeros, rather than triggering a referent retrieval procedure, signal that no retrieval is needed, since there is no alternative referent choice.

20 This refers to Grice's cooperative principle. For a lucid treatment of this principle see also Clark [1996: 140-153]. Halliday & Hasan express the same view, saying that "we insist on interpreting any passage as text if there is the remotest possibility of doing so" [1976: 23].

occasionally will there occur minor cases of pending identification, where the reader-listener will have to go ahead within the text in order to be able to identify the reference of a subject retrospectively, as in (10):

(10) И оумоливь 0 (ABR) некого оужикоу (rec) свокго посла 0 (ABR) тамо, да извтсто оувтсть 0 (ABR?/REC?) w нки и знаменакть 0 (REC) м'Ьсто гд^ ходить 0 (MAR). [BS 6r]

He asked a certain relative and sent him there to find out about her and to indicate the place where she was residing.

Judging from the contents and structure of the first part of the sentence both ABR and REC are grammatically possible referents for the expression извтсто оувтсть 'that he may find out the truth', and using the zero form— or, for that, the infinitive as in our English translation—leaves the identity of who actually is meant to be the subject reference, open. Knowledge of the narrative situation does not help either: it is the relative who is being sent to inquire about Mary, but only for the ultimate goal of Abraham's finding out the truth about her. There would be in fact no way of coming to know the subject reference, were the expression извтсто оувтсть 'to find out about her' to stand by itself. However, the expression is conjoined with знаменакть мтсто 'that he may indicate the place' in a coordinate construction, and the reference of conjoined verbs must be the same. For знаменакть мтсто the identity of the referent can be decided on the basis of the narrative situation. It is the relative and not Abraham who goes on the reconnaissance mission, and so it can only be him who is in a position to identify Mary's whereabouts. Thus, only on reading-listening further will the reader-listener come to know the identity of the referent. This is, of course, a minor case, which, even if the identity would be left pending, could in no way impede the comprehension of the story as a whole.

Since identification of referents, as discussed here, operates on the verbatim level of utterances, the resolution of pending identity may be supposed to be limited by the short-term memory buffer [Givon 1992: 44-45]. It has been shown that verbatim expressions do not pass into long-term memory on hearing or reading them, and the capacity of this buffer is said to comprise no more than one or two clauses. Pending identity within the confines of a sentence should therefore not cause any disruptive effects on text processing, whereas beyond the sentence limit it is technically irresolvable, unless retrieval of verbatim expressions outside the memory buffer is safeguarded, as is the case only with written texts being read privately by an individual instead of being recited to an audience.

It turns out that the available formal contrast between zero forms and nouns is not employed systematically to solve the problems of referential

conflict. There is only one case throughout AQM, where NR could be argued to solve a referential conflict:

(11a) Сь ним (abr) же бд'Ьше и поаше 0 (MAR) и акоже самь (ABR) вьздрьжаше се, такожде и wна (MAR) присно пр'Ьдьсп'Ьющи вь чрьноризьчьст'Ьмь мбраз'Ь тьщаше се сьврьшити вьсе д'Ьтели. (11b) Ибо множицею блажени (ABR) сь сльзами молаше се бгоу (god) w нки (mar), да би им^ла 0 (MAR) мысль свою кь нкмоу (god) и да не привежет се 0 (MAR) никакоиже земльн'Ьи вещи, понеже бо б'Ь Щь ек (BRA) им'Ьник доволно wставиль. [BS 1v-2r]

She used to keep vigil and sing with him and, just as he practiced abstinence, so too she strove to attain to the perfection of his deeds, constantly excelling in her hermit's life. The blessed man would frequently supplicate God on her behalf, asking that she direct her mind to him (God) and not get entangled by any of the world's attractions, now that her father had left her ample money.

By formally opposing ABR as самъ to MAR as она by way of a contrasting comparison in (11a), both ABR and MAR acquire an equal activation status, so that both qualify equally for the main clause subject position in (11b). On the face of it, NR here seems to serve the task of helping the reader-listener decide who in fact fills in the subject slot. But then there is a clear indexical cue (w нки), which is sufficient to determine the identity of the subject referent, so there is no true referential conflict. Perhaps, the case at hand ought to be treated as a special case of coping with formally induced equal activation, deriving from contrasting procedures as in (11a), which will form a barrier for referential choice by forcing the reader to choose between the two opposing referents put on offer by means of the contrastive comparison. A somewhat similar effect of formally induced equal activation might also be argued to be brought about by absolute constructions which exhibit a partial referential overlap with the subject of the main clause:

(12) И вьшедьшима же има (ABR/MAR), вид'Ь 0 (ABR) wдрь высокь настлань. (10r) When both had entered, he saw a large bed made up.

Though absolute constructions usually are not involved in referencing procedures, in this special case the partial overlap of the referent of the absolute construction and the main clause may be expected to induce an interpretation of the subject of the absolute construction as a set of alternatives to choose from. In this case, however, the formally induced equal choice does not work as a barrier. Referential choice is in this particular case determined by pro-tagonisthood. ABR is a continuing prominent topic of the preceding sentences, and zero reference is chosen accordingly. The equal choice does not block access to the preceding discourse. Though it is tempting to treat (11a)-(11b)

in terms of either a true or a formally induced referential conflict, on closer consideration it turns out to be a rather murky case.

What then is the function of the contrast between nominal and zero forms, if any.? There is one quite obvious context where a noun rather than a zero form ought to be used, and this is the introduction of a new topic-to-be into the text. This will more often than not take place in subject position, but it does not need to. Of course, the number of personal agents that become topical in AQM is very limited indeed, so this specific function cannot account for all cases of nominal reference. Apart from that, introducing new topics cannot count as NR.

The functions NR serves can best be gleaned from the role of the two most prominent topical figures of the story, viz. the title figures ABR and MAR. For ABR only 14 cases of NR could be identified, and for MAR the number is even lower with only 4 cases. The quite specific domain of adjacency pairs, where NR is employed to identify the summoner and the respondent at the beginning of a conversation, accounts for 4 of all 18 cases.

In two cases there occurs a shift from the narrative to the comment level as in the short comparison in (13) embedded into the ongoing plot:

(13) aKo^e 6o nacTHpb, erga w6p,fem,eTb wbto noru6meK, cb pagocTHro Bb3bMeTb k Ha paMo cbok, TaKo h 6^a^eHH (ABRhyp) pagyK ce ughme. [BS 13v]

Like a shepherd, once he has found his sheep that has gone astray, takes it happily on his shoulders, the saint went happy at heart.

The shift to the comment level implies also a shift of the role of Abraham. Though ABR figures in (13) also as an acting person, he becomes at the same time the object of contemplation for the narrator, who changes his footing from reporting events to contextualizing them and their participants into a higher moral-metaphysical framework. Accordingly, the protagonists will change their role to objects of metaphysical reflection. Change of footing from narration to commentary causes also a break in sequentiality of the reported events and can be argued to affect conjoinability as defined by Li & Thompson [1979: 329-330], thereby triggering the use of a marked form of resumption. We assume that this change of footing, which interrupts the narrative flow, triggers NR even with prominent topics that dominate the scene into which the comment is embedded. A somewhat similar case is presented by resumption by means of descriptive items, as is the case in (14):

(14) B'fecTa gbHH gbBa Hmbaa no ombCTBHH 6.na®eHbia (mar), bi Ha Bugfe 0 (ABR) bh^^hhk, gbB^ .I'hT'h, H^e cbTBopu KpoM'fe Kro (abr) gbrnepbmu Kro (MAR). [USP 301r]

Now the two days that passed since her departure and in which he saw the visions, became two years, which his niece spent abroad.

Note that MAR is already nominally introduced as an adnominal adjunct in the first part of the sentence, so there is no need for NR at the end of it, where zero anaphora instead of the nominal expression дъщерьши юго would suffice. In fact, the nominal form is used here to convey a short comment. By using дъщерьши юго instead of just applying the default nominal reference блаженнаш (эталонная форма 'labelling form' see below), the particular responsibility that Abraham assumes for Mary as his relative is being stressed. In this case, as in many others, nominal resumption is used to provide a comment and control the moral interpretation of the events by highlighting specific aspects or properties of the referent.21 Commentary functions involving change of footing or descriptive items account for another four cases.

On the face of it remoteness seems to come into play in the use of resumptive стры юе in (15b) for ABR, who temporarily has ceded the role of main protagonist on the scene to MAR:

(15a) Прилежно же чрьноризьчьствоваше 0 (MAR) по всемоу чрьньчьствоу кго.

(15b) Веселшше же се стры (ABR) ке, виде ю (mar) доброе прилежанык, сльзы и см'Ьренноую моудрость, безьмльвык и кротость и любовь, юже им'Ьше 0 (MAR) кь боу.

(15c) к. л^т чрьноризьчьствова 0 (MAR) сь нимь шко агница чстаа, и голоубыца нескврьннаа. [BS 2v]

Willingly she trained herself in all the excellent ways of her blessed uncle. Her uncle delighted to see her fine intention, her tears and her humble wisdom, her quiet and gentle nature and her love, which she had for God. For twenty years did she lead a hermit's life with him, being like a chaste lamb and a spotless dove.

But then, at this point ABR is not as remote as would make his reactivation necessary. He ceded his lead role to MAR only shortly before, though, of course, the aboutness of the scene has changed with MAR taking over. Instead of Abraham's solicitousness about Mary's becoming a true hermit, it is now Mary's eagerness to follow her uncle that is being highlighted. It appears that zero referencing is reserved to the person who is the subject of the theme or aboutness of the passage at hand. Within this context, ABR is assigned the role of supplying a comment on this overarching theme. He is in fact not even an actor on the scene, but assumes the role of giving a commentary aside. Thus, ABR is not just demoted to the position of a secondary protagonist, he temporarily loses his protagonisthood as such. We assume that zero reference is indeed reserved to actual protagonists, so that non-protagonists would have to be marked nominally.

21 This use of nominal anaphora is reminiscent but not identical to the phenomenon of metaphorical anaphors as treated by Helge Skirl [2007].

Neither remoteness nor referential conflict with SER/DIA could be invoked to account for NR by блаженыи in (16c). ABR has been introduced by the same nominal expression блажены only in (16a), and as a matter of fact, he is the only person present at the location of the narrative scene reported in (16a-c):

(16a) Прилоучившоу же се семоу еи (mar), вид'Ь вид'Ьник блажены (ABR), страшнаа змиа (ser) велика з'Ьло, мрьзька ©бразомь, шоумеща (ser) кр'Ьпостию, изьшедшаа (ser) wx м^ста свокго и пришедша (ser) до кклик его (abr).

(16b) И wбр,feть 0 (SER) голоубицю (col), погльти 0 (SER) ю (col), и пакы вьзврати 0 (SER) се вь м'Ьсто свок.

(16c) Вьспреноув же се wx сна блаженыи (ABR) и мскрьб'Ь 0 (ABR) з'Ьло и плака 0 (ABR) се горцЬ, еда оубо гонкник вьздвигнеть сотона (DIA) на стюю црьковь и многык wставить 0 (DIA) wx в'Ьры, или разореник боудеть црквы. [BS 5r-v]

After this had happened to her, he (Abraham) had a dream: he beheld a huge serpent disgusting to look at and hissing in a fearsome way. On leaving its lair it came toward his cell. It found there a dove, swallowed it up, and went back to its lair. When the blessed man awoke, he was much troubled and started weeping bitterly. "Perhaps Satan is going to stir up persecution against the Holy Church and alienate many from the faith, or maybe there will be schisms in the church (of God)."

The proper interpretation of the scene would certainly not be impeded, if блаженыи in (16c) would be replaced by zero. What induces the use of NR in this case, remains a bit of a riddle. It appears to have to do with the switch from the level of reported narrative reality to the reality of the dream and then back to the narrative, which would license rather than enforce the use of NR in this case. The switch of levels could then be treated as a subepisodic shift with (16c) marking an event boundary. In the light of the somewhat similar constellation of figures in (17a-b) this interpretation ought, however, to be taken with a grain of salt:

(17a) И пакы по двою дьнию вид'Ь 0 (ABR) тогожде змиа (ser) изьшедша (ser) wx м'Ьста свокго, и пришедша (ser) вь хизиноу кь нкмоу (abr), подложивша (ser) же под ноз'Ь блаженаго (abr) главоу свою, и расЬдша се (ser), и мбр'Ьтени быти голоубици (BS 5v) wnou (COL) въ чрпвп егш (ser).

(17b) простьр' же блженыи (ABR) роукоу и приеть 0 (ABR) ю живоу не имоущу скврьны. [NBKM 299, 116r]

Then once again, two days later, he saw the same serpent leave its lair and come toward his house, whereupon it placed its head beneath the blessed man's feet; the serpent was then ripped open, and there, safe and sound was found the dove in its belly. The blessed man stretched out his hand and took his relative, who was unharmed.

As in (16c) the resumptive nominal 6naMemi in (17b) is preceded by a scene in which SER/DIA adopts a leading role, though within a dream sequence. This time, however, 6naMerni in (17b) forms part of the dream sequence, so that no boundary effect can be claimed here. Perhaps, a shift of the major acting protagonist's role from SER/DIA to ABR must be assumed to account for NR in both cases: thus, dramatic role patterns rather than remoteness or referential conflict appear to control the referential choice.

The only case in which remoteness, i.e. the need to reactivate an otherwise prominent topic that happens to have been absent from the narrative for a longer stretch of text, applies appears to be (16a), which reintroduces ABR as a main protagonist after his effectively being absent from the scene as an active participant for a significant stretch of text.

Kibrik [2011: 38-39] observes that NR will, as a rule, make use of one fixed expression (etalonnaja forma) throughout any one text instead of employing a wide range of different descriptive labels. This appears also to be true for NR in AQM, though it must be pointed out here that this is probably not a universal discourse feature. There are, in fact, culturally defined literary genres which will require constantly changing reference labels for poetic reasons, like the stylistic feature of kenning in Old Norse alliterative Skaldic poetry [Fidjestol 1974]. Looking once again at our main protagonist Abraham, we find the following nominal identifiers for him. After having been introduced as 6naMeHHuu aepaMU in the title22 and once again as aepaMiu at the outset of the story, he is being uniformly referred to as 6naMeHHuu in 10 out of a total of 14 cases of nominal subject resumption. In two cases he is referred to by the expression cmpuuIn both cases, Abraham is interacting with Mary, so that the expression seizes upon the opportunity to stress their specific blood relationship. The remaining two cases consist of complex NPs which are clearly meant to provide an evaluative comment on the doings of Abraham reported in the respective sentences (caMb mb dueHuu MyMb, UMe e .H. nmmb npbHbUb-cmea ceокго He ebKycu xnn>6a 'that very same miraculous man who during 50 years of his monk's life did not eat (even) a piece of bread'). An interesting feature which holds also for the other main protagonist, his niece Mary, is that NR never makes use of the proper name as identifier. Further research ought

22 It should be noted that proper names, being semantically void, do not establish the identity of a referent by themselves as would regular lexical entities. They are informative about the referent to the extent that only they are anchored somehow to the wider context of the narrative on their first appearance in the text. This can be and often is done by some kind of introductory description [Marslen-Wilson et al. 1982]. In the case of hagiography, rubrication of the saint as the main protagonist in the title serves as an anchor for the reference of his proper name. As the protagonists of hagiographic texts all belong to the same class of people (saints), the title 'saint' may be considered as functionally fully equivalent to the proper name of the saint, so that both proper name and title could be used interchangeably.

to sort out whether this is possibly a general stylistic feature of the hagiog-raphic genre, or at least of part of its tradition.

3.2. Episodes as boundaries for referencing

It is commonly held that the demarcation of units, such as episode boundaries, typically involves explicit referential forms [Hinds 1977; Levinsohn 1978; Longacre 1979; Georgakopoulou 1997: 114; Kibrik 2011: 405]. The basic defining requirement for paragraphs is thematic unity [Longacre 1979: 118120]. Narrative episodes, which can be seen as genre-specific instantiations of the more generic concept of paragraph, require in addition to thematic unity: unity of time, place and main protagonists. Episode transitions will therefore usually involve a change in at least one of these parameter settings [Longacre 1979: 118]. The start of a new episode within the narrative, though it does not involve recalibration of the reference coordinates or change of footing, as is the case with direct speech and commentary asides, requires reintroduction of topics by means of reactualisation strategies. It is not so much that episode breaks would make former topics remote, but rather that every episode has to define the scene of action including its protagonists all over again, because it cannot be taken for granted that protagonists from preceding episodes will automatically participate in the new episode. Explicit referencing (mostly nominal) appears to be required at the first mention of an already established prominent topic. In AQM, however, it appears that nominal resumption at episode boundaries is restricted to topic entities entering the scene for the first time within the narrative or after a period of absence from the preceding epi-sode(s), as in (18), when ABR re-enters the scene in the Dream Episode after the completion of the Seduction Episode, where he played no part in:

(18=16a) Прилоучившоу же се семоу еи, вид'Ь вид'Ьник блажены (ABR), страшнаа змиа велика з'Ьло, мрьзька ©бразомь, шоумеща кр'Ьпостию, изьшедшаа wx м'Ьста свокго и пришедша до кклик его. (5r)

After this had happened to her, the saint had a dream: he beheld a huge serpent disgusting to look at and hissing in a fearsome way. On leaving its lair it came toward his cell.

Both types can be qualified either as new topics-to-be or as remote topics, which would both require nominal (re)introduction, anyway. If the figure in question happens to be the most prominent topic entity in the immediately preceding episode, zero anaphora is used instead to introduce a protagonist, as at the beginning of the Reconnaissance-episode in (19d)23:

23 The use of zero anaphora across episode boundaries is found in two other cases within the text; in all three episodes ABR is the only acting figure. These findings are at odds with claims that episode boundaries will enforce full reference, "even if the referent was recently mentioned" [Kibrik 2011: 405]. For a somewhat similar statement about

(19a) Въздвиже 0 (ABR) же гласъ свои, глагола 0 (ABR) съ сльзами: He raised his voice in tears, and said:

(19b) "Съпасе вьсего мира, Христе, възврати агницю твою Марию въ оградоу жизньноую, да не сънидеть старость моа съ печалию въ адъ.

(19c) Не пр'Ьзьри молкниа мокго, господи, нъ посьли благодать твою въскор'Ь, да изьметь ю изъ оустъ змиквъ." [USP 301в]

O Christ Savior of the world, return the lamb Mary to the fold of your flock that I may not go down to hell full of grief in my old age.

Do not turn aside my request, Lord; rather, send your grace at once and deliver her from the serpent's mouth.

EPISODE BOUNDARY

(19d) Б'Ьста же дьни дъва, ишьла по ошьствии блаженыа, въ на же вид'Ь 0 (ABR) вид'Ьник, дъв'Ь же л'Ьт'Ь, иже сътвори кром'Ь кго (abr) дъщерьши кго. [USP 301г] Now the two days that had passed since her departure and in which he saw the visions, became two years which his niece spent abroad.

In this particular case, the choice of zero anaphora may have been additionally favored under the impact of the same subject resumption rule which holds for resumption following intervening direct speech sequences. ABR appears at this stage of the sequence of episodical events as the only continuing topic entity, anyway, so zero anaphora would appear a natural choice for referring back to him. It looks like the rules for introducing protagonists into an episode are very much the same as hold for prominent vs. remote/new topics in general, which would ultimately mean that episode boundaries do not define a specific subset. An important qualification must, however, be placed on this possible claim. The unusually high degree of protagonisthood of ABR, which can be seen as a characteristic feature of hagiographic narratives, may account for the permissibility of zero referencing across episode boundaries in this case, so that specific rules may still be applicable to less prominent figures.

paragraph boundaries, see Longacre [1979: 118]: "Lack of back reference is indicative, therefore, of a paragraph boundary", which would exclude zero anaphora, but also some types of absolute constructions which mark an episode break by means of a discourse topic anaphora referring back to the events of the preceding episode as being completed. See also Ariel [1991: 448]. See, however, Levinsohn who states for the Inga language of Colombia: "Between episodes in which the cast and roles of major participants vis-à-vis each other remain constant {...} reference to the participant(s) involved is zero" [1978: 95, see again 97]. An obvious way out in order to save this general claim would be that the episode boundaries in question were assigned incorrectly. In fact, the only protagonist within these episodes is ABR, so that a case could be made for treating the events in question as part of just one episode. Apart from the risk of circularity, however, which this revision of our structure assignment would run, the episodic structure is strongly suggested by a time lapse of two years between the Dream- and the Reconnaissance-Episode and the insertion of a longer praising comment between the Reconnaissance- and the Rescue-Episode.

4. Conclusion

The findings of our study of the CSL translation of the Story of Abraham and his niece Mary (AQM) are by and large corroborated by Eckhoff & Meyer's [2011] quantitative study into null subjects in OCS, performed on the PROIEL corpus. The general tendency of undercoding subjects appears thus not so much a specific feature of hagiographic narratives or even more narrowly of AQM, but seems to reflect a quality of OCS/CSL itself. This statement should, however, come with a warning not to misconstrue it as specifying a feature of a language system. In saying that CSL displays a strong tendency to undercode referents in subject position, we treat CSL not so much as an autonomous linguistic system, but as a cultural tool which - in the way it is applied - reflects the culturally determined attitudes of its users with respect to what they consider proper communicative behaviour.

In our study we have opted for an experiential approach which not only brings into focus the reader-listener and his or her efforts to make sense of what he or she reads or hears, but which at the same time frames participant tracking as part of the more general cultural technique of reading. We hope to have also been able to show that undercoding practices of communication, though, of course, asking a lot from the reading interpreter, can be taken very far indeed without any serious threat of a major communicative breakdown. This is possibly the point where genre comes into play. Hagiography takes protagonisthood to extremes by putting an exceptionally strong focus on the figure of the saint. Much of the text of a saints' life will stage the saint and his thoughts and actions as the one and only figure which deserves full attention, leaving little room for true interaction (especially more complex situations involving more than two interactants). As a matter of fact, the strong personal focus of hagiographic writing is directly reflected in the distribution of participants over the available subject slots within the narrative portions of the text. 94, i.e. 44.6% out of a total of 210 subject slots, are occupied by ABR. AQM, however, is exceptional among hagiographic writing in allowing for a complimentary protagonist alongside the saint, i.e. MAR who occupies 75 (35.6%) of all subject positions. Another 6.8% go to non-animate nontopical subjects, so that only 13% remain for minor protagonists. It should be quite obvious how this genre-specific personal focus makes referential conflict almost a non-issue for subject-tracking. The impact of genre properties may also be gleaned from the treatment of topic continuity across episode boundaries. It could be shown that universalist claims of episode boundaries requiring marked forms of topic resumption do not hold for saints in saints' lives, due to their particularly marked protagonisthood.

Undercoding, as attested in participant tracking in AQM, ultimately shows language to be no more than an admittedly highly useful and potent

expedient to communication. However, it is certainly not a kind of grammar machine that provides for the technically exact transmission of thought, by providing, e.g., for the exact coding and decoding of participants.24 Though CSL confirms universally observed features of dealing with the construction of cohesion through tracking continuing topics throughout a text, by going through the experience of a modern reader trying to come to grips with an individual text composed in a remote past there emerges a distinct feeling for the specific manner in which medieval Slavic text production dealt with the needs and expectations of its audience, which starkly contrasts with those of modern Western European readers. Clearly, CSL tends to Kibrik's 'egoistic' end as it comes to taking care of the reader, but the term itself - apart from its undesirable connotations - somehow seems to miss an important point. Un-dercoding as a common practice is not about some kind of personal deficiency or even cognitive impairment, it is rather about the culturally agreed 'proper' style and manner of telling a story in writing. This makes techniques of participant tracking a feature of culture more than anything else.

List of actants appearing in the examples

ABR Abraham

COL pure dove of Abraham's dream vision

COR heart

DIA devil

GOD god

HOS innkeeper

MAR Mary

MON monk who seduces Mary

REC scout sent by Abraham to find Mary

SER evil serpent of Abraham's dream vision

0 zero

Capital letters mark actants in subject position, small letters those in none-subject position.

Bibliography Source manuscripts

BS

Bdinski Sbornik, Ghent University Library ms. 408, date given in colophon 1360, but probably a later copy from the first decades of the 15th century. Facsimile edition: Dujcev 1972.

Uspenskij Sbornik, GIM Uspenskij 4/1063, end 12th-beginning 13th century. Edition: Kotkov 1971.

24 See Roy Harris' [1981] critique of what he terms the telementation fallacy.

NBKM 299

Paraenesis, Bulgarian National Library ms. 299(2), anno 1592. Description: Conev 1910, vol. 1, 228-233.

Literature

Ariel 1991

Ariel M., The function of accessibility in a theory of grammar, Journal of Pragmatics, 5 (16), 1991, 443-463. Bielfeldt 1961

Bielfeldt H. H., Altslawische Grammatik, Halle an der Saale, 1961. Brock, Harvey 1987

Brock S. P., Harvey S. A., Holy Women of the Syrian Orient, Berkeley, Los Angeles, London, 1987. Chafe 1982

Chafe W. L., Integration and involvement in speaking, writing, and oral literature, Spoken and Written Language: Exploring Orality and Literacy, D. Tannen, ed. Norwood NJ, 1982, 35-53. Clark 1996

Clark H. H., Using Language, Cambridge, 1996. Conev 1910

Conev B., Опис наръкописите и старопечатните книги на Народната библиотека в София, 1, Sofia, 1910. Dujcev 1972

Dujcev I. S., ed., Bdinski Zbornik. Ghent Slavonic Ms 408, A.D. 1360. Facsimile edition, London, 1972. Eckhoff, Meyer 2011

Eckhoff H., Meyer R., Conditions on null subjects in Old Church Slavonic, a contrastive view. Paper presented at the 20th International Conference on Historical Linguistics, Osaka, July 25, 2011.

Fidjestol 1974

Fidjestol B., Kenningsystemet: Forsek pä ein lingvistisk analyse, Maal og Minne, 1974, 5-50. Gaelens 2017

Gaelens J., Reference and topic continuity in Old Church Slavonic narrative texts. A case study based on the Bdinski Sbornik. MA thesis, University of Ghent, 2017.

Georgakopoulou 1997

Georgakopoulou A., Narrative Performances. A study of Modern Greek storytelling, AmsterdamPhiladelphia, 1997. Givon 1983

Givon T., Topic Continuity in Discourse: An Introduction, Topic Continuity in Discourse. A quantitative cross-language study, T. Givon, ed., Amsterdam-Philadelphia, 1983, 1-41. - 1992

Givon T., The grammar of referential coherence as mental processing instructions, Linguistics, 1 (30), 1992, 5-55. Gordon et al. 1993

Gordon P. C., Grosz B. J., Gilliom L.A., Pronouns, names and the centering of attention in discourse, Cognitive Science, 3 (17), 1993, 311-347. Grimes 1975

Grimes J. E., The Thread of Discourse, The Hague-Paris, 1975. Halliday, Hasan 1976

Halliday M. A. K., Hasan R., Cohesion in English, London, 1976.

Harris 1981

Harris R., The Language Myth, London, 1981. Haspelmath 2013

Haspelmath M., Argument indexing: a conceptual framework for the syntactic status of bound person forms, Languages across Boundaries. Studies in Memory of Anna Siewierska. D. Bakker, M. Haspelmath, eds., Berlin, Boston, 2013, 197-226.

Hinds 1977

Hinds J., Paragraph structure and pronominalization, Papers in Linguistics, 10, 1977, 77-99. --1983

Hinds J., Topic continuity in Japanese, Topic Continuity in Discourse. A quantitative cross-language study, T. Givon, ed., Amsterdam, Philadelphia, 1983, 43-93.

Kibrik 1987

Кибрик А. А., Механизмы устранения референциального конфликта, Моделирование языковой деятельности в интеллектуальных системах, А. Е. Кибрик, А. С. Нариньяни, ред., Москва, 1987, 128-145.

--2011

Kibrik A. A., Reference in Discourse, Oxford, 2011. Kotkov 1971

Успенский сборник XII-XIII вв., под ред. С. И. Коткова, Moscow, 1971. Levinsohn 1978

Levinsohn S. H., Participant reference in Inga narrative discourse, Anaphora in discourse, J. Hinds, ed., Edmonton, 1978, 69-135. Leskien 1969

Leskien A., Handbuch der altbulgarischen (altkirchenslavischen) Sprache, 9th ed., Heidelberg, 1969. Li, Thompson 1979

Li Ch. N., Thompson S. A., Third-person pronouns and zero-anaphora in Chinese discourse, Syntax and Semantics, 12: Discourse and Syntax, T. Givon, ed., New York, 1979, 311-335. Longacre 1979

Longacre R. E., The paragraph as a grammatical unit, Syntax and Semantics, 12: Discourse and Syntax, T. Givon, ed., New York, 1979, 115-134.

Lunt 2001

Lunt H. G., Old Church Slavonic Grammar, 7th ed., Berlin, New York, 2001. Marslen-Wilson et al. 1982

Marslen-Wilson W., Levy E., Tyler L. K., Producing interpretable discourse: the establishment and maintenance of reference, Speech, Place and Action. Studies in Deixis and Related Topics, R. J. Jarvella, W. Klein, eds., Chichester, New York, 1982, 339-378. Mithun 1996

Mithun M., Overview of general characteristics, Handbook of North American Indians, 17: Languages, I. Goddard, ed., Washington D.C., 1996, 137-157. Olson 1977

Olson D. R., From utterance to text: The bias of language in speech and writing, Harvard Educational Review, 47, 1977, 257-281. Payne 1997

Payne Th. E., DescribingMorphosyntax, Cambridge, 1997. Schank, Abelson 1977

Schank R., Abelson R., Scripts, goals, plans and understandings, Hillsdale NJ, 1977.

Scharpé, Vyncke 1973

Scharpé J. L., Vyncke F., Bdinskizbornik. An Old-Slavonic Menologium of Women Saints (Ghent University Library Ms. 408, A.D. 1360), Bruges, 1973. Skirl 2007

Skirl H., Metaphorical anaphors. A phenomenon of the semantics-pragmatics interface, Anaphors in Text. Cognitive, formal and applied approaches to anaphoric reference, M. Schwarz-Friesel, M. Consten, M. Knees, eds., Amsterdam, Philadelphia, 2007, 103-119. Vaillant 1964

Vaillant A., Manuel du vieux slave, 2nd ed., Paris, 1964. van Dijk, Kintsch 1983

Dijk T. A. van, Kintsch W., Strategies of discourse comprehension, New York, 1983. Vecerka et al. 1993

Vecerka R., Keller F., Weiher E., Altkirchenslavische (altbulgarische) Syntax, 2: Die innere Satzstruktur, Freiburg im Breisgau, 1993. Zwaan 2003

Zwaan R. A., The immersed experiencer: toward an embodied theory of language comprehension, The Psychology of Learning and Motivation, 44, 2003, 35-62.

References

Ariel M., The function of accessibility in a theory of grammar, Journal of Pragmatics, 5 (16), 1991, 443463.

Bielfeldt H. H., Altslawische Grammatik, Halle an der Saale, 1961.

Brock S. P., Harvey S. A., Holy Women of the Syrian Orient, Berkeley, Los Angeles, London, 1987.

Chafe W. L., Integration and involvement in speaking, writing, and oral literature, Spoken and Written Language: Exploring Orality and Literacy, D. Tannen, ed., Norwood NJ, 1982, 35-53.

Clark H. H., Using Language, Cambridge, 1996.

Dijk T. A. van, Kintsch W., Strategies of discourse comprehension, New York, 1983.

Dujcev I. S., ed., Bdinski Zbornik. Ghent Slavonic Ms 408, A.D. 1360. Facs. ed., London, 1972.

Fidjestol B., Kenningsystemet: Forsok pä ein lingvistisk analyse, Maal og Minne, 1974, 5-50.

Georgakopoulou A., Narrative Performances. A study of Modern Greek storytelling, AmsterdamPhiladelphia, 1997.

Givon T., The grammar of referential coherence as mental processing instructions, Linguistics, 1 (30) 1992, 5-55.

Givon T., Topic Continuity in Discourse: An Introduction, Topic Continuity in Discourse. A quantitative cross-language study, T. Givon, ed., Amsterdam, Philadelphia, 1983, 1-41.

Gordon P. C., Grosz B. J., Gilliom L.A., Pronouns, names and the centering of attention in discourse, Cognitive Science 17, 3, 1993, 311-347.

Grimes J. E., The Thread of Discourse, The Hague-Paris, 1975.

Halliday M. A. K., Hasan R., Cohesion in English, London, 1976.

Harris R., The Language Myth, London, 1981.

Haspelmath M., Argument indexing: a conceptual framework for the syntactic status of bound person forms, Languages across Boundaries. Studies in Memory of Anna Siewierska. D. Bakker, M. Haspelmath, eds., Berlin, Boston, 2013, 197-226.

Hinds J., Paragraph structure and pronominali-zation, Papers in Linguistics, 10, 1977, 77-99.

Hinds J., Topic continuity in Japanese, Topic Continuity in Discourse. A quantitative cross-language study, T. Givon, ed., Amsterdam, Philadelphia, 1983, 43-93.

Kibrik A. A., Mechanisms of Referential Conflict Removal, Modelirovanie iazykovoi deiatel'nosti v in-tellektual'nykh sistemakh, A. E. Kibrik, A. S. Nari-nyani eds., Moscow, 1987, 128-145.

Kibrik A. A., Reference in Discourse. Oxford, 2011.

Kotkov S. I., ed., Uspenskii sbornik 12th-13th c., Moscow, 1971.

Leskien A., Handbuch der altbulgarischen (altkir-chenslavischen) Sprache, 9th ed. Heidelberg, 1969.

Levinsohn S. H., Participant reference in Inga narrative discourse, Anaphora in discourse, J. Hinds, ed., Edmonton, 1978, 69-135.

Li Ch. N., Thompson S. A., Third-person pronouns and zero-anaphora in Chinese discourse, Syntax and Semantics, 12: Discourse and Syntax, T. Givon, ed., New York, 1979, 311-335.

Longacre R. E., The paragraph as a grammatical unit, Syntax and Semantics, 12: Discourse and Syntax, T. Givon, ed., New York, 1979, 115-134.

Lunt H. G., Old Church Slavonic Grammar, 7th ed., Berlin, New York, 2001.

Marslen-Wilson W., Levy E., Tyler L. K., Producing interpretable discourse: the establishment and maintenance of reference, Speech, Place and Action. Studies in Deixis and Related Topics, R. J. Jar-vella, W. Klein, eds., Chichester, New York, 1982, 339-378.

Mithun M., Overview of general characteristics, Handbook of North American Indians, 17: Languages, I. Goddard, ed., Washington D.C., 1996, 137-157.

Olson D. R., From utterance to text: The bias of language in speech and writing, Harvard Educational Review, 47, 1977, 257-281.

Payne Th. E., Describing Morphosyntax, Cambridge, 1997.

Schank R., Abelson R., Scripts, goals, plans and understandings, Hillsdale NJ, 1977.

Scharpé J. L., Vyncke F., Bdinski zbornik. An Old-Slavonic Menologium of Women Saints (Ghent University Library Ms. 408, A.D. 1360), Bruges, 1973.

Skirl H., Metaphorical anaphors. A phenomenon of the semantics-pragmatics interface, Anaphors in Text. Cognitive, formal and applied approaches to anaphoric reference, M. Schwarz-Friesel, M. Consten, M. Knees, eds. Amsterdam, Philadelphia, 2007, 103-119.

Vaillant A., Manuel du vieux slave, 2nd ed., Paris, 1964.

Vecerka R., Keller F., Weiher E., Altkirchensla-vische (altbulgarische) Syntax, 2: Die innere Satzstruktur, Freiburg im Breisgau, 1993.

Zwaan R. A., The immersed experiencer: toward an embodied theory of language comprehension, The Psychology of Learning and Motivation, 44, 2003, 35-62.

Dieter Stern, Professor, Doctor

Universiteit Gent

Fakulteit Letteren en Wijsbegeerte

Vakgroep Talen en Culturen

Blandijnberg 2

9000 Gent

Belgie / Belgium

[email protected]

Received March 29, 2019