36
MARION KRAUSE, CHRISTIAN SAPPOK, OLGA YOKOYAMA ACCENTUAL PROMINENCE IN A RUSSIAN DIALECT TEXT: AN EXPERIMENTAL STUDY 0. INTRODUCTION Our investigation presents a methodological and theoretical approach to Russian dialect intonation. This is not a totally new field of research (cf. Брызгунова 1977; Пауфошима 1983 and 1989; Люблинская, Саппок 2002). We believe that programs for data collection should take intonation into account in addition to traditional grammatical categories (cf. Касаткин 1991, 9). This recommendation corresponds with a widely accepted notion that the speaker of a dialect or of a language variety in general reveals specific dialectical traits not least in this area. However, any attempt to describe these intonational differences is faced with the following questions: Does the language variety in question have a separate intonational system in its own right, a different set of intonational rules creating contours and corresponding meanings that are not found in other language varieties or, as dialect is involved, in the standard language? Or is it the case that the intonational system remains the same and that there are other factors determining speakers’ differing behaviour? As a consequence of this latter assumption, we would expect to find alternative principles of textual organization favoured in the variety or dialect we are confronted with. To test this hypothesis we created two versions of a spoken dialect text, a narration stemming from the north Russian region of Archangel’sk – the original spoken version and an orthographically normalized written version – and used the two versions as stimuli in a perceptual experiment. The results justify the assumption that alternative principles of accent assignment can be applied to one and the same basic text and that the resulting texts remain acceptable in both versions and correspond to different means of information processing based on different preferences in dialect and standard. 1. THE TASK In the programmatic title of his article ‘Accent is predictable (if you’re a mind reader)’ Bolinger (1972) emphasizes that the prediction of what comes next in the text and what the place of accent will be are two different undertakings. The first has to do with the objective sequence of events to be reported, while the second has to do with subjective intentions in the mind of the speaker. The analysis of the Russian Linguistics 27: 251 – 286, 2003. 2003 Kluwer Academic Publishers. Printed in the Netherlands.

Accentual Prominence in a Russian Dialect Text: An Experimental Study

Embed Size (px)

Citation preview

MARION KRAUSE, CHRISTIAN SAPPOK, OLGA YOKOYAMA

ACCENTUAL PROMINENCE IN A RUSSIAN DIALECT TEXT:AN EXPERIMENTAL STUDY

0. INTRODUCTION

Our investigation presents a methodological and theoretical approach to Russiandialect intonation. This is not a totally new field of research (cf. Брызгунова1977; Пауфошима 1983 and 1989; Люблинская, Саппок 2002). We believethat programs for data collection should take intonation into account in addition totraditional grammatical categories (cf. Касаткин 1991, 9). This recommendationcorresponds with a widely accepted notion that the speaker of a dialect or of alanguage variety in general reveals specific dialectical traits not least in this area.

However, any attempt to describe these intonational differences is faced withthe following questions: Does the language variety in question have a separateintonational system in its own right, a different set of intonational rules creatingcontours and corresponding meanings that are not found in other languagevarieties or, as dialect is involved, in the standard language? Or is it the casethat the intonational system remains the same and that there are other factorsdetermining speakers’ differing behaviour? As a consequence of this latterassumption, we would expect to find alternative principles of textual organizationfavoured in the variety or dialect we are confronted with.

To test this hypothesis we created two versions of a spoken dialect text, anarration stemming from the north Russian region of Archangel’sk – the originalspoken version and an orthographically normalized written version – and usedthe two versions as stimuli in a perceptual experiment. The results justify theassumption that alternative principles of accent assignment can be applied to oneand the same basic text and that the resulting texts remain acceptable in bothversions and correspond to different means of information processing based ondifferent preferences in dialect and standard.

1. THE TASK

In the programmatic title of his article ‘Accent is predictable (if you’re a mindreader)’ Bolinger (1972) emphasizes that the prediction of what comes next in thetext and what the place of accent will be are two different undertakings. The firsthas to do with the objective sequence of events to be reported, while the secondhas to do with subjective intentions in the mind of the speaker. The analysis of the

Russian Linguistics 27: 251 – 286, 2003. 2003 Kluwer Academic Publishers. Printed in the Netherlands.

252 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

latter mechanism will leave the events and their sequence out of consideration; itwill, instead, focus on the intimate processes of the speaker’s mind as he expresseshis individual attitudes and evaluative activities through the assignment of accentto elements of the text.

Bolinger (1972) posits two layers of structuring any process of communication.One gives rise to the textual form reflected in the written version of the utterance,dictated primarily by circumstances outside the speaker’s influence. The otherlayer reflects the speaker’s freedom to choose accentual prominence accordingto his evaluation of events and their textual appearance, place, and relevance.According to Bolinger, the linguist analysing accentuation has to gain access tothis aspect of the speaker’s mind.

But in our opinion, it is not only the linguist who is confronted with the taskof interpreting what is in other peoples’ minds. The speaker does something verysimilar with the mind of his hearer. This is, to put it simply, the approach de-veloped within the framework of the Transactional Discourse Model (henceforthTDM, Yokoyama 1986): The speaker assesses the state of his addressee’s mind,his knowledge system, i. e., the knowledge set as a virtual pool to be activated andthe subset that is actually activated, whether by preceding contexts or by other fac-tors. In the TDM the latter subset is given the label “matters of current concern”.An interpretation of the current concern in the hearer’s mind can be based onactual, objective observation; on the other hand, it can be assumed and positedby the speaker, a process that is called “imposition”. In any case, a decision hasto be made, even at risk of mis-assessment, because the final shape of the utter-ance, including its prosody, has to reflect those decisions. The speaker has made,as soon as he finished his utterance, a certain set of assumptions or, in terms ofBolinger (1972), predictions as to the mind of his hearer. These assumptions area substantial part of his message.

This roughly outlined discourse situation is of general validity, and it also re-mains valid in the case of cross-variational communication, i. e., in a situationwhere the speaker and hearer have a common language but employ different va-rieties or different dialects of this common language. We also have to incorporateinto this type of communication what has been said above about the differenttypes of predictions being made by the speaker about the listener and by the lin-guist about the author of the text. We are always confronted with a basic textand the intonational features given to it in actual, i. e., audible performance. Thisgives rise to our crucial questions concerning intonation in general and accent as-signment as one of the constituent operations within the intonational shaping ofthe utterance in particular: Is the sense of a deviant prosodic structure based on anon-conforming set of intonational rules, or is it based on a different type of in-formation processing, with different predictions about the state of the addressee’smind? Is the accent assignment one that would seem to be normal between inter-

PROMINENCE IN A RUSSIAN DIALECT TEXT 253

locutors who share the same language variety but that would be deviant from theperspective of a listener coming from outside the situation and from outside of theoriginal language variety?

To test the hypothesis of concurrent principles of textual organization asopposed to concurrent systems of intonational rules we had to find the appropriatematerial: an authentic utterance consisting of a sequence of sub-strings that wouldallow for alternative choices concerning the assignment of accent. This mightseem to be difficult if not impossible. As a rule, we are ready to accept alternativeaccentuation within one and the same string of words. We expect, on theother hand, differing, mutually exclusive positions of this string within the text.Исаченко (1967) was one of the pioneers formulating the correlation betweenword order, accent position, and textual coherence. In the further developmentmodels concerning the information structure of sentences it has been convincinglyshown that there are cases of prominence assignment which are not exclusivelydetermined by context and which are, as a consequence, open to speakers’individual and local decisions (cf. Molnár 1993).

A text which contains several instances which are open to alternative accentualsolutions could perfectly well be used as a platform to study the mechanisms ofaccent assignment in general as well as the role of this assignment in the case ofcross-variational and cross-dialectal communication. It would be possible to testto what extent the original solutions of the dialect speaker are in conflict withthe solutions a standard speaker would prefer when presented with a normalizedwritten version of these original utterances. In addition, it would be possible totest to what extent the original accentuation is accessible and seems acceptableto listeners who are not dialect speakers. In carrying out these two types ofexperimental procedures we can gain some insight into one of the intonationaloperations of a dialect speaker through the ears of a standard speaker.

2. THE EXPERIMENTS

Having in mind this general task of observing prominence assignment in com-parable contexts we conducted two experiments. Our subjects were speakers ofstandard Russian who could be expected to be not totally unfamiliar with Russiandialects. The first experiment was designed to test subjects’ predictive ability; itsaim was to bring to light the expectation of someone presented with a writtentext and intending to give it an audible shape. In accordance with this aim weused stimuli in written orthographic form to be pronounced with an appropriateprominence contour. In the second experiment subjects were presented with audi-ble pieces of the same text and were asked to make a choice as to the place of themost prominent word, this time on the basis of the heard stimuli. Subjects againhad to assign the place of prominence by marking this place in the written version,

254 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

this time not on the basis of prediction, but by assessing the heard version of thewritten form.

We made three basic assumptions as to the nature of auditory knowledge andthe metalinguistic competence of our subjects:

1. Subjects reading a fragment of text and understanding its place within theoverall text’s thematic, informational, and semantic context would be ableto consider the different options of accent assignment and to check themagainst their interpretation of the fragment.

2. Subjects hearing and interpreting a spoken piece of speech would be able tomake attempts to specify the most prominent unit, using the acoustic inputas well as their own experience of how prominence is most usually signalled.

3. These two different mechanisms are not only part of the subjects’ perceptualcompetence, but also of their overall competence in organizing communica-tion. We thus expected the following: The textual frame staying as it is therehas to be a certain amount of freedom in assigning alternative prominencecontours and favouring one of them. This set of alternatives is an integralpart of a language user’s overall competence in assigning accentual promi-nence.

This set of assumptions led us to our choice of methods. We used text fragmentsof parallel appearance in both experiment I (predictive) and experiment II(perceptual). We picked out cases exhibiting a high degree of agreement withinthe group of subjects in each experiment. We next asked whether the placeof accent assignment was convergent or divergent in these pieces of text. Theinstances of divergence gave us a primary hint as to the range of alternativesand freedom we posited at the beginning. Although we confine our analysisto these instances, we must not forget the differences between the roles of thewritten form of the stimuli in the two types of experiments. In experiment Iit represents stretches of standard speech that are candidates for future acts ofacoustic realization; in experiment II by following orthography rather than thedialect it originated from, the written text is far from the actual audible signal andfunctions merely as a text to be read easily and to be used to mark the place ofperceived prominence.

There is one further source of possible conflict between the two versions. Thespoken version comes from a dialect speaker while the judgments are made bystandard speakers. One might suspect that we are comparing standard prominenceassignment with dialect prominence assignment as heard through the ear of astandard speaker. If this is true, one system is misused as it is being comparedand evaluated by a non-authentic competence. We are aware of this potentialconflict, but we have no other alternative. We would, if logically following thisline of argumentation, have no way of getting analytic access to dialect prosodyat all. We would thus have to admit that there might be differing and mutually

PROMINENCE IN A RUSSIAN DIALECT TEXT 255

non-accessible intonation systems. But we will not, in fact, exclude an alternativeview: That there is a common system of intonation and prominence assignmentused by different groups within the language community and that differences inprominence assignment do not arise from differences in the underlying system ofrules, but from variants in prosodic performance and from preferences regardingthe use of different contours chosen from the same set of well-formed contours.We will, at least partly, also check this hypothesis of a common system on thebasis of convergent instances of prominence assignment found in our data.

We now come to our working hypothesis that determines our access to theproblematic field of variation and prosody. Prominence assignment exhibits acertain degree of freedom, even in cases where the context remains the same. Atthe same time, the process of assignment is governed by a strict set of principlesthat structure the information that is to be transferred through a text. A text’swritten form, addressed to the reader, i. e., to our subjects as readers, should reveala high degree of consent. The written stimuli should give rise to a highly uniformdistribution of responses, but the set of alternative solutions would not be excludedeither by grammar or by context.

This hypothesis justifies our assumption that the judgements stimulated byexperiment II, based on the authentic utterances, would not be classified as ‘right’or ‘wrong’, but as having a legitimate place in a spectrum of alternative choices.All of them, we may conclude, are rule governed, not context governed. We expectthat a model such as the TDM would be able to describe the exact nature of themechanisms involved. What we are interested in is casting some light into thismechanism of structuring information that is to be transferred through a text. Inextending the level of tolerance of acceptable prominence contours, we hope tocontribute to the development of an extended model: Intonation in the field oflanguage varieties should no longer be seen as a matter of alternative systems, butas a field of alternative principles of information structuring.

2.1. The stimuli and their source

The stimuli we used in our experiments were collected from a text published inКасаткина (1991) in both acoustic and written (i. e., transcribed) form. We cutout fragments of the original acoustic versions while preserving the coherence ofthe text that would be necessary for the functioning of the experimental procedureand the interpretation of the results.

The procedure of segmenting the dialect narrative had to take into account itsspoken nature and the perceptual processing of the resulting segments. In orderfor the fragments to serve as objects of perception while not showing defects ofthe cutting procedure two criteria had to be maintained: The fragments had to befully interpretable as natural spoken units; and they had to be units corresponding

256 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

to units of the natural utterance process, so-called phrasing units. In the case ofconflict between the two criteria we favoured the first, which gives some of thestimuli a format one would not expect if oriented only towards the written version.The fact that perceptual segmentation and syntactical segmentation do not alwayscoincide is well known (cf. Светозарова 1988a; Скорикова 1997).

The experiment is based on the assumption that the interpretation of the text, theperceptual evaluation of its constituent parts and the solution of the experimentaltask is not determined exclusively by the text in its written form. This expectationhas its justification in the nature of a spoken utterance. The details will bediscussed later when the instances of divergence have evolved as results of theexperiment. It will suffice, for the moment, to have a look at the type of textand the situation in which it was uttered. The situation is typical of the so-called“narrative interview” (Schuetze 1983) with evident autobiographical orientation:The dialectologist visits the dialect speaker and asks him to tell something abouthis former life. As a consequence, the result will be communication with acharacteristic multi-level structure. On the one hand it is oriented towards alistener who comes from far away, without knowing anything about the personhe is talking to and bringing nothing with him but his interest. On the other hand,the text depicts the informant’s process of remembering, i. e., his sorting out ofinstances that are stored in his memory and put together in a kind of discoursewith just himself. Bearing this in mind, we expect different types of coherenceand different coexisting ways of structuring the narrative flow. This expectationwill be justified by the nature of our results, as we will see below.

2.2. The experimental procedures

25 subjects took part in the experiments, all adults with Russian mother tongueliving in St. Petersburg. They had no explicit knowledge of the phonetics ofRussian or of Russian dialectology. The experiments were carried out using aportable computer.i

In experiment I the stimuli were presented individually to each subject on thecomputer screen in written form in normalized orthography, that is, without anyhint as to the dialectal source. A full version of the text is given in Appendix I.After a short intermediate pause experiment II followed, in which the subjectwas presented with the original spoken form of the same fragments in additionto the written version. In both sessions the fragments used as stimuli werepresented in the same order as in the original narrative. Each written text fragmentdisplayed on the screen had a space under it where the subject could place thesign for the prominent portion, predicted or really heard. The software used isdescribed in Knipschild, Sappok (1991). The acoustic version of the stimuluscould be repeated, the repetition triggered by the subject himself, as long as the

PROMINENCE IN A RUSSIAN DIALECT TEXT 257

decision process was going on. After this decision, it was impossible to return toa preceding fragment, and it was also impossible to hear a fragment in advance.

After the presentation of a stimulus the participants were required to select aword as being the best place of prominence. The instructions did not excludethe assignment of prominence to more than one word within one and the samestimulus. This multiple assignment could not be excluded in view of the fact thatthere were fragments with more than one syntactical unit that could potentiallybe accented which could not be further segmented for auditory reasons (seethe description of the cutting procedure above). Independent of these technicalrespects it is not possible to exclude the possibility of more than one prominentword within one and the same phrasing unit. This phenomenon is describedby Fant, Kruckenberg (1999); similar observations for Russian have been madeby Светозарова (1988b) and Кодзасов (1996). Prominence can be, as aconsequence, either binary (i. e., on one word as opposed to the rest of the unit)or in steps (i. e., the prominence of one element not excluding the occurrence ofother prominent elements within the same unit).

The first step in dealing with the results was to collect the answers, to countthe number of times each word of the text had been chosen. The results of thisprocedure can be seen in columns 3 and 4 of Table 6 (see Appendix II). The nextstep was to assign each word a prominence degree using a scale from 0 to V (seeTable 1 for an overview).

Table 1. The scale of the relative prominence degrees

If a word was judged as being prominent by almost all of the subjects (81%or more), its prominence was given a V on this scale. It turned out to be optimalto use the phonetic word – that is, a word with its pro- and enclitics (henceforthwe’ll call it “word” to avoid complicated formulations) – as a unit of prominence.This is in accordance with the subjects’ behaviour of never designating prefixesor postposed particles such as -to as prominent.

258 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

As a result of this procedure, each stimulus received a prominence profileas depicted in columns 5 and 6 of Table 6. There are two different profilescorresponding to subjects’ evaluations in experiments I and II. The profilesrepresent the result of our summing up all of the subjects’ decisions regardingthe location of prominence. Subjects were not given the opportunity to make adecision as to the degree of prominence. The value V does not mean that anevaluation had an extra high prominence; it merely means that the word wasdesignated as prominent by almost all of the subjects.

The statistical evaluation of the responses from the written stimuli as comparedto the responses in the perceptual experiment was performed using McNemar’smethod (see Lohse et al. 1982, 123 ff.) for two correlated subsets. The degree ofsignificance of the differences showing up between the two experiments is givenin column 7 of Table 6. With few exceptions, the cases of divergent prominenceassignment (2 steps on the scale) coincide with a statistical significance in thedifference between responses.

2.3. The results in general

Table 6 (see Appendix II) is an overall presentation of the results in the followingorder: Columns 1 and 2 contain the written versions of the stimuli together withthe name of the corresponding file. The percentage of times that each phoneticword was marked as prominent in the written version is noted in Column 3 andin the audible version in Column 4. Columns 5 and 6 show the same results in asimplified scale. Column 7 shows the significance of the differences between thetwo types of stimuli, either being non-significant (n.s.), of medium significance(m.s.) or having a high degree of significance (h.s.).

Before we come to a detailed analysis of the results we will make somegeneral remarks concerning the distribution of responses. As a rule, the agreementbetween the subjects was rather high; the profiles show a clear-cut maximumrepresenting the one favoured decision. This is to be seen in the distribution ofwords according to the number of times they were chosen as being prominent.Table 2 shows the results for the written stimuli as compared to the results for theaural stimuli.

As expected, the number of words which were assigned prominence by amajority of subjects (reaching level IV or V on the prominence scale) is relativelysmall. A comparable distribution is described in the work of Elsner (1999) andthe work of Streefkerk et al. (1999). For each stimulus that means that there isa preferred profile. But there are also stimuli with a tendency to yield a diffusepicture and, accordingly, a flat profile, either of the stimulus as a whole or of partsof it, that is, of one of its constituent phrases.

PROMINENCE IN A RUSSIAN DIALECT TEXT 259

Table 2. Distribution of prominence degrees (0–V) on phonetic words (in %): outcomes of thepredictive and the perceptual experiments. The lower chart presents the same results graphically

We may sort out three types of “diffuse” profiles. The first group shows aweakly structured profile in both versions, as written material as well as inacoustic form. The profile is similar for both versions in 0007b, 0015b-I and0022b-II. The second group is diffuse in one mode and well structured in the other.0012-cI, 0015b-II, 0022b-I are well structured in their written form; in contrast,0025 gained a more profiled set of answers as an aural stimulus. The third groupcan be classed as diffuse, but with a divergent distribution in the two differentmodes (0022b-I).

What is the overall picture when we compare the profiles resulting fromexperiment I with the results of experiment II? In cases where we have roughlythe same type of profile in both versions, there is a tendency towards moreexplicitness in experiment II: aural input makes the decision more unequivocaland definite. But it is not always the case that a stimulus becomes more profiledor more definite as to the localization of its prominence after being heard. Profilesof one and the same token resulting from experiments I and II may differ in shape,and sometimes it is the case that the written version is more definite than the heardversion. This can be explained in three ways: (a) the spoken realization of the

260 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

original stimulus can be diffuse, not showing a sharp contour; this can be seenin the case of 0007b; (b) there is a relatively sharp contour of prominence in theoriginal stimulus, but as its place does not correspond to the subjects’ predictions,it does not become as sharp as it could have become; (c) the decision to class aword as prominent might be complicated by dialect features such as okan’e (cf.Пауфошима 1983, 64).

In the next section we will concentrate our analysis on cases with a clear-cutdifference between the two versions. These examples are of excellent relevancefor our model of alternative prominence assignment within the text and for a con-sideration of the consequences of this phenomenon for the adequate descriptionof dialect intonation.

3. THE ANALYSIS OF THE DIVERGENT CASES

3.1. The data

The most striking among the cases of divergence between the results of experi-ment I and experiment II, the judgments related to the written and the aural stim-uli, are those where the relative prominence of certain elements of an utteranceis reversed in the two experiments. The number of subjects evaluating the promi-nence of words in token 0007a, for example, are presented in Table 3 (see § 2.2.for a discussion of the problem of phonetic words).

Table 3. Distribution of answers to stimulus 007a

In the written experiment, the majority of the subjects (76%), when evaluatingthe second sentence (words (3)–(5)), predicted the prominence of the adverbialphrase (5), and no one considered the subject (3) prominent. However, six subjects(24%) assessed (4) to be the prominent word. The results of the aural evaluationwere different: 72% perceived the prominence of the predicate verb, while theadverbial phrase (5) was judged to be prominent by only 40 percent of thesubjects.ii

PROMINENCE IN A RUSSIAN DIALECT TEXT 261

When the numerical results are translated into a scale of 0–5, the contrastsamong judgments are sharpened, and the same results for token 0007a then appearas in Table 4.

Table 4. The distribution of judgments given to stimulus 0007a in a simplified scale

In the following sections 3.1–3.3, we will refer primarily to these simplifiedresults, which will be represented graphically. The following is the graph forthe same token 0007a: the results of the written experiment for each word areconnected by a solid line, while those of the aural experiment are connected bya dotted line. Concerning the areas where the lines cross we will refer to theendpoints of the solid line as A and B and to the endpoints of the dotted line as Cand D.

The reversed relative prominence of words (4) and (5) of stimulus 0007a inthe two experiments is expressed in the graph by the crossing of AB and CD(further discussion see below; the correspondence between numbers in the graphsand words in the text can be found in Column 2 of Table 6 in Appendix II). Inthe written experiment, the uniformity of the auditors’ judgments is considerablygreater with regard to the prominence of (5); in the aural experiment, on thecontrary, the majority agreed on the prominence of (4). Note that the differenced and the concentration of prominence judgments between the two experimentson both sides of the intersection equals 2, i. e., d = C − A = B − D = 4 − 2.

262 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

We will consider below only cases where d ≥ 2, at least on one side of thecrossing. In the first sentence on pal, the difference d of the evaluation of wordsequals 3, but such cases will not be considered: despite the magnitude, theyreveal only the difference in the number of auditors judging word (2) to beprominent, all of whom nevertheless judged word (2) to be more prominent thanword (1) in both the written and the aural experiments. The crossing of the lines,then, reveals a relative difference between the auditors’ projected evaluation ofaccentual prominence and their evaluation of the prominence they perceived inthe informant’s oral rendition of the text. It is precisely this sort of difference thatconcerns us here.

The total percentage of the marked words in the written experiment equals104%, which means that one of the auditors (4%) assumed two words to beprominent, while the rest of them marked only one word each. 64% assumed thatsuch a prominent word was the adverbial phrase (4). Approximately 1/3 (32%)assigned prominence to (1), posle. No one assumed prominence of the subject,and at least one (4%) of the three auditors who marked (3) considered it to be themost prominent word of all four.iii

In their evaluation of the aural text the majority of the auditors (84%) found thepredicate prominent. The place adverb (4) was perceived to be prominent only by12% of the subjects, while posle was not marked by anyone.

In the first sentence (on pal), no one considered the subject (1) to be prominentin either experiment. As for the predicate (2), it was assumed to be prominent

PROMINENCE IN A RUSSIAN DIALECT TEXT 263

by the majority (84%) in the written experiment, although only 40 percentthought it to be prominent in the rendition by the dialectal narrator. In the secondsentence (ja pobežala v derevnju), the assessment of the relative prominence ofthe predicate and the adverbial phrase resembles the one in 0005b. The majorityof the auditors (76%) judged, based on the written text, that the adverb (5)iv

would be prominent, while 24% assigned prominence to the predicate (4). Noone considered the pronominal subject to be prominent. In evaluating the auraltext, however, the majority (72%) identified the prominence of the predicate (4),while some of the same subjects also marked the adverbial phrase (5), the totalproportion of its markings being 40%.

In the first sentence, the majority of the subjects (84%) indicated that theywould assign prominence to the predicate (2), while none of them consideredthe subject (1) a likely candidate for accentual prominence. In their evaluation ofthe aural text, on the other hand, the results varied: almost half (48%) perceivedprominence on the subject (1), while the other half was divided between thosewho found the predicate (2) prominent (24%), and those who identified noprominence either on the subject or the predicate (28%).

The clitics da and ved’ were presented to the subjects in the written experimentas two separate words, and were both correctly identified by all subjects asaccentless. For the purpose of the graph, therefore, they were incorporated intothe preceding or the following fully accented words Valerko and pomrët,v giventhe impossibility of determining whether the auditors considered the unaccentedparticle sequence da ved’ to be proclitics or enclitics. The choice was irrelevantfor the graph configuration, as the values remain 2 and 5 for words (5) and (6),respectively, in the written experiment, and 4 and 4 for the same words in the auralexperiment. Focusing on the relative evaluation of words (5) and (6) in this way,it is easy to see that for these two words, in the written experiment, the majorityof the subjects (96%) assumed that the predicate (6) would be prominent, whileonly 36% (including some of those who also marked the predicate) assumed theprominence of the subject (5). In the evaluation of the aural text, the perception of

264 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

the prominence of (5) and (6) was comparably well-represented (80% and 72%,respectively), nearly half of the auditors marking both words.vi

Most of the subjects (92%), based on the written experiment, assumed theprominence of the imperative verb, four of them (16%) also marking the vocative,the total proportion of those marking it being 24%. The narrator’s rendition wasevaluated with less unanimity: the judgment went either for the vocative (56%) orfor the verb (44%), evidently with no overlap.

The only crossing considered here is that between the results for (4) and (5).vii

The second sentence in 0022b, where this crossing is found, produced results thatsplit, and were in complementary distribution. Nearly half (44%) considered thesubject (4) to be the prominent word, with 28% preferring to assign prominenceto the direct object (3), and another 24% settling on the utterance-final predicate.One auditor (4%) did not mark any of the three words. This failure to identifyany word as prominent rose to 12 percent in the evaluation of the aural data,although a slightly more convincing majority (56%) now perceived prominenceon the predicate, while prominence markings regarding (3) and (4) fell to 20%and 12%, respectively.

In both the written and the aural experiment, the total markings for this tokenwas 104%, indicating that all but one auditor marked only one of the two wordsas prominent. In the written experiment, for the majority (72%) this prominentword was the nominal element (1). In the aural experiment, on the other hand, itwas the predicate (2) with an even greater level of unanimity (92%).

PROMINENCE IN A RUSSIAN DIALECT TEXT 265

The results for this token were virtually identical to the previous token. Onlyone (4%) of the auditors marked both words, while the rest selected one or theother word for prominence. In the written experiment, the majority (72%) againfavored the nominal element (1). In the aural experiment, on the contrary, 96%perceived the predicate to be prominent (2).

In the written version of the test, a majority (92%) assigned prominence inthis token to the verb (1), evidently considering it to be an independent sentence.A majority also marked the subject (3) and the vocative (7) as prominent (80%and 76%, respectively). Notably, in the second sentence it was the subject ratherthan the predicate (12%) that most auditors assumed to be prominent; 8% of theauditors did not pass any prominence judgments regarding the second sentence.The range of answers was greater for the third sentence: while 76% predicted thatthe locative would be prominent, a sizable 52% also considered the pronominalsubject to be accentuated, and 28% marked the predicate as well. The total

266 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

markings in the third sentence added up to 156%, which means that more thanhalf of the auditors in this three-word sentence marked two words each asprominent.viii

The range was narrower when the aural test was evaluated. Only 20% ofthe auditors perceived the first verb (1) to be prominent. In both the secondand the third sentence, i. e., in the sequences (2)–(4) and (5)–(7), only 16%heard prominence on the subjects, while the predicates, on the contrary, werejudged to be prominent by 48% and 60% for the second and the third sentences,respectively. The greatest convergence of opinion regarding prominence (80%)was found with respect to the locative (7).

A total of 25 markings were made for this token, showing that it was only themost prominent words that received them. In the written experiment, the majority(80%) predicted prominence on the nominal element, the direct object. In theperceptual experiment, on the contrary, prominence was heard on the nominalelement only by 8%, while the remaining 92% marked the verb.

3.2. The analysis of the divergent tokens

The significant discrepancies between the expectation of prominence by thespeakers of the literary language and their perception of prominence in theaural data as rendered by the dialectal informant fall into three groups. The firstgroup consists of cases in which the majority of the auditors predicted utterance-final prominence, which is characteristic of narrative intonation in the distantcužoj mode (section 3.2.1).ix Cases in which the majority of auditors assumedprominence of certain nominal items but heard it realized on the predicatesconstitute the second group (section 3.2.2). The third group is analyzable ascases in which the majority assumed different syntagmatic divisions or syntacticrelationship in their prognosis when compared to the decisions they heard in thedialectal informant’s narration (section 3.2.3). We consider each group separatelybelow.

PROMINENCE IN A RUSSIAN DIALECT TEXT 267

3.2.1. Failed expectations of narrative intonation in the cužoj mode

These were tokens 0012b and 0012c, as described above. The expectations of theauditors were the same in all three cases of intersecting graphs: the assessmentby the majority was that the last element in the sentence was to be prominent.The less prominent of the two elements in these bi-syntagmatic sentences wasjudged by most subjects to be the referential expressions ona and Valerko in0012b and Griška in 0012c. Such an assessment is to be expected of speakersof the Russian literary language, such as the subjects in our experiments, giventhe usual deaccentuation of the referential items which the speaker estimates tobe found in the interlocutor’s set of current concern at the moment preceding theutterance.x The subjects heard a different intonational solution in the renditionby the dialectal narrator. In all three cases the majority noted prominence to beimplemented precisely on the referential expressions ona, Valerko, and Griška.xi

The range of opinion differences in all three cases, however, was considerable:in the aural experiments the predicates were found to be prominent only byslightly more subjects than the referential expressions. Evidently, the subjectswere ambivalent about the accentual characteristics of these three cases in whichthe dialectal narrator produced intonation quite different from what the speakersof the literary language would have produced.

We cannot say for sure what exactly the subjects reacted to when evaluatingtokens 0012b and 0012c. The two tokens in question, however, stand out in severalways that must have played a role. One of the factors that must have contributedin particular to the difficulty of evaluating the intonation of the dialectal narratorwas her agitated narration. As can be seen from the Fo curve in the graphspresented in Appendix III, the pitch level in the vicinity of 0012a sharply riseson the stressed syllable of ona in 0012b-1 and 0012b-2 and remains quite high(300–325 Hz) throughout, until the middle of 0012c. Such levels of Fo are notfound anywhere else in the entire narrative, where all other peaks, including othercases of quoted speech, do not exceed 250 Hz, with the exception of only fourother peaks reaching 290 Hz. Marija Ivanovna’s tearful imprecation renderedby the narrator as direct speech is completed in 0012c, the first word of which,the vocative Griška, is still very high at 300 Hz. It is only after this vocativethat the pitch falls to the more usual level of 200 Hz. Moreover, token 0012bincludes an interjection (oj) and two discourse particles. These features point to ahighly marked discourse, not only lexically and intonationally, but also in termsof its discourse characteristics (direct speech, vocative) and meaning in general(pleading, invoking a life-and-death situation). In fact, the intonational contourthat very likely reflects the difficulty the speakers of the literary language hadwhen faced with it appears to be a lament-like rhetorical intonational formula,featuring an LH tone implemented on the first stressed syllable and an HL

268 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

tone implemented on the last syllable; in the stretch of segmental materialbetween these two contour tones, each full lexical item is accentually prominent,while maintaining an H-phrase accent without a perceptible declination of thebaseline.xii The speakers of the literary language did not expect such a formulahere, yet they seem to have recognized its features at least in part, perhaps becauseit is not entirely alien to Colloquial Russian. Although such a formula has notbeen described previously, it may in fact share something with IK-5 identifiedby Брызгунова (1980). Thus we suggest that in their assessment the auditorsassumed the narrative intonation of the distant cužoj mode, while in the renditionof the dialectal narrator they heard a vaguely familiar but unexpected lament-likeformula, which contributed to the wide range of their evaluations.

3.2.2. Failed expectations of prominence of the nominal elements in the svojmode

These are the cases of graph crossings in 0022b, 0022c, 0023a, 0029, and 0031. Inthese cases, the majority of the auditors projected prominence on the referentialitems Marija, sani, bereg, mašiny, vse, and operaciju. In the narration, on thecontrary, they identified these items as deaccented, while at the same time theyperceived their predicates as prominent. Token 0022b stands apart from the rest ofthese cases in that the discrepancy between the written and the aural experimentswas caused by insufficient information made available to the subjects. In theabridged written experiment, Marija (Ivanovna) appears for the first time, whilein the aural experiment, the narrator was operating with her full context, in whichMarija (Ivanovna) had appeared earlier in her capacity as the narrator’s chiefhelper in the task of saving her son. Having already mentioned Marija (Ivanovna),the narrator naturally placed her referential knowledge in Ca∩Cb, and this ledto its deaccentuation in token 0022b. Although the subjects themselves had noreason to deaccentuate Marija, most of them still accurately perceived the lack ofprominence of this referential item in the dialectal narration. In the cases of graphcrossings in 0022c, 0023a, 0029, and 0031, most subjects assigned prominenceto the referential items and deaccentuated the predicates that follow them. Thiswas a highly natural solution: faced with written data in which these previouslyunmentioned referential items appear before their respective predicates, they didthe obvious, giving prominence to these non-anaphoric items. This inevitablyforced them to assume the intonation of the svoj mode, thereby shifting out ofthe previous cužoj mode.

What most of them heard in the aural text was the prominence of the predicatesand the corresponding deaccentuation of the referential items. Unlike the case ofMarija in 0022b, where the narrator had an objective reason to treat the referentialexpression as anaphoric, the narrator’s choice to deaccentuate the referential

PROMINENCE IN A RUSSIAN DIALECT TEXT 269

expression while giving prominence to the predicates could not have beenobjectively motivated on anaphoric grounds. What motivated the prominence ofthe predicates was the narrator’s unilateral, impositional assumption that her ownmental vision of the events was shared by the hearer(s). Evidently, the narrator wasreliving the scene which these referential items were part of, and the sequence ofthe events that came into being was in the center of the narrator’s attention. Theutterances under consideration address the questions she ascribes to her hearer(s)as she addresses them: “Did they put him on the cart (or not)?”, “Did they takehim to the shore (or not)?”, “Were the cars already waiting (or not)?”, “Did theyperform the operation (or not)?”.

In the TDM (Yokoyama 1986), this egocentric act of assuming that whatconcerns the speaker also concerns the hearer, which then leads to the form ofan utterance that reflects this assumption, is called “imposition”. In the utteranceunder consideration, the imposition was evidently accepted by the auditors whenthey heard it, even though in their own assessment of the written text they did notresort to an imposition. The acceptance of the imposition by the auditors musthave been induced by their acceptance of the narrative “contract” assumed by thenarrator. Unlike the narrator of folktales, who begins by introducing the existentialand the referential knowledge of the characters (cf. Жил-был царь. У негобыло три сына.) and only after these have been introduced proceeds to thetransfer of propositional knowledge (Старший сын был . . . etc.), our dialectalnarrator of a personal reminiscence operates with a different contract, one closerto that of a literary narrator, who does not hesitate to impose onto the readera propositional question like Что случилось в доме Облонских? withoutfirst introducing the Oblonskys (cf. the opening of Tolstoy’s Anna Karenina). Inour case, the dialectal narrator takes the liberty of not introducing the actantswho have appeared to her mental vision, and imposes propositional disjunctionsabout the predicates involving these actants. The difference between the dialectalnarrator’s approach and the approach of the speakers of the literary language boilsdown to the following: the latter do not undertake imposition, presupposing amore objective discourse situation and a less involved narrator who takes a moreobjective picture of the cognitive situation as his point of departure; the formerdoes not hesitate to impose what concerns her own self at a given moment.

3.2.3. Diverging syntagmatic division and/or sentence structure

Unlike the cases considered in 3.2.1 and 3.2.2, where the difference between theresults of the written and the aural experiments was based on different intonationalsolutions to a different assessment of the discourse situation, the crossings of thegraphs in 0005b and 0007a, we suggest, are caused by differences in the syntactic

270 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

structures assumed by the speakers of the literary language and by the dialectalnarrator.

In their prognosis for token 0005b, 32% of the auditors indicated prominenceon the utterance-initial posle. Twice as many assumed prominence on the unitna ulicu. Such a distribution may suggest that those who marked posle must haveassumed that this word constituted a rising utterance-initial syntagm, as wouldbe normal for distant neutral intonation, provided that the word is interpretedadverbially as ‘later’. The remaining two-thirds of the subjects, on the other hand,may have reacted to the falling intonation of the utterance-final syntagm, alsonormal for the neutral distant intonation pattern (Yokoyama 2001). In the auralexperiment, most subjects (84%) heard the predicate to be prominent, and noone perceived prominence on posle. This suggests that in the rendition of thedialectal narrator they heard a monosyntagmatic phrase posle on ušël na ulicu.If this majority was correct in assuming that the narrator had produced only onesyntagm here, with prominence realized on the predicate and a complete lackof it on posle, 0005b ceases to be an independent sentence. Instead, we mustnow analyze it is a subordinate clause with posle functioning as a subordinatingconjunction posle togo kak, followed by the main clause 0005c.

Ultimately this analysis hinges on the acceptability of such a sub-standardconjunction-like meaning of posle in this dialect, which remains to be confirmed.The narrative as a whole, however, does contain other cases of conjunctionlesssubordination (cf. 0007a and 0029), which offer indirect support for such aninterpretation. In token 0029, which was already discussed above in section 3.2.2,this concerns the evaluation of the first word doleteli, which was not discussedin section 3.2.2, as no crossing of the graphs was observed after this word.The difference in the prominence assignment of this word itself between thewritten and the aural experiments is, however, quite pronounced, and it suggests adifference in the syntactic structures assigned to this word in the two experiments.For their own projection, 92% of the subjects marked this word as prominent,evidently assuming that it constitutes an independent syntagm and sentence,followed by the next sentence tam už mašiny gotovy. In contrast to this, in therendition of the narrator only 20% perceived it to be prominent. Apparently, inthe spot where they themselves would have placed a “period”, the majority hearda “comma” in the speech of the dialectal informant. A “comma” would of coursemean that doleteli is subordinated to the following sentence, meaning ‘w h e nthey got there, there were already cars waiting’, despite the absence of an overtconjunction ‘when’.

It is our contention that the crossing of the graphs in 0007a is also caused bya discrepancy in the structure assumed in the two versions of the experiment.In their assessment, a majority of the subjects appear to have chosen essentiallytwo clauses juxtaposed to one another in a coordinated construction: On – pal,

PROMINENCE IN A RUSSIAN DIALECT TEXT 271

(a) ja pobežala v derevnju. In a distant mode, the clause-final elements paland v derevnju would in such a case carry a falling HL contour tone, easilyinterpretable as accentual prominence, just as the majority had predicted. Indeed,in the first clause prominence of the predicate was projected by 84%, and in thesecond clause, prominence of the adverbial phrase by 76%. Notably, the presencein the adverbial phrase of the particle -to, marking a previously mentioned item,did not prevent 76% of the auditors from assessing prominence on the wordcontaining it: normally, a sentence-final position, combined with HL tone, shouldmean that the phrase is “new”, that is, not assumed to be within the scope of thespeaker’s attention. Such a problematic solution could only have been motivated,we suggest, by an overriding contrastive interpretation of the two sentencesconstituting the sequence 0007a.

In the aural experiment, on the other hand, the prominence of the predicate inthe first clause was perceived only by 40%, as was that of the adverbial phrase inthe second clause. Instead, it was the predicate that 72% found to be prominent inthe second clause. Moreover, since it was the predicate of the second clause thatwas perceived by the majority as the more prominent of the two clause predicates,for this majority, at least, the second clause must have been perceived to bethe main one, with the first clause being subordinated to it, the meaning being:‘When he fell, I started to run’. According to such an interpretation, v derevnju-toacquires the less problematic reading of an afterthought, consistent with theparticle -to. Thus, the results of the written experiment, we suggest, were borneout by a contrastive coordination reading of the two clauses, as was natural toassume for a narrative by a distant narrator with a corresponding distant narrativeintonation. The prominence the auditors perceived in the aural experiment, onthe other hand, leads to a subordinate reading in a complex conjunctionlessconstruction.

In this discussion of tokens 0005b, 0007a, and 0029, we base our analysis ofthe syntactic structure on the intonational reading of these utterances as perceivedby the subjects. The crucial role of rhythmic and intonational means in theformation of conjunctionless complex sentences has already been mentioned byПешковский (1928). The existence of such conjunctionless subordination in theRussian literary language must have contributed to the evaluation of prominencein the dialectal rendition of these utterances.

3.3. Conclusions

The proposed analysis of the divergences between the prominence in the writtensentences projected by the speakers of the Russian literary language and theirperception of prominence in the spoken rendition of the same utterances by adialectal narrator suggests several fundamental differences between the discourse

272 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

strategies of these two types of Russian native speakers, as well as between theirsyntactically-determined assignment of prominence. The majority of the speakersof the literary language first and foremost projected an orderly organization ofa narrative text. They did not assume any impositional relationship with theaddressee, instead preferring the distant cužoj mode whenever possible, avoidingafterthoughts or conjunctionless subordination. Nor did they assume a particularlydramatic rendering of direct speech incorporated in the narrative.

The subjects were able to detect in the speech of the dialectal narrator allthe deviations from their own projected prominence, despite the differencesin prominence assignment and in discourse strategies. They took notice ofaccentual weakening of conjunctionless subordinate clauses and afterthoughts,as well as of the deaccentuation of the referential expressions whose knowledgewas imposed by the narrator onto the addressees, along with the impositionof the relevant propositional disjunctions. They also reacted to, if only byindicating their confusion, the lament-like formula used by the narrator in directspeech.

The character of the projected intonation pointing to the distant cužoj modeassumed by the speakers of the literary language and revealed in their projectionsis notable in and of itself. Also interesting is the fact that they were able toaccept all the dialectal deviations from their own solutions. While their orientationtowards the distant addressee is consistent with the great interlocutor distanceprevailing in literary narrative “contracts”, their ability to adjust to the dialectalnarrative intonation and syntax suggests that the dialectal features in question arenot entirely alien to the speakers of the standard language and must be stored intheir passive linguistic competence. This analysis represents the first attempt totackle the novel area of dialectal intonology. While focusing first and foremoston the intonational aspect of interdialect communication, it at the same timecontributes to our understanding of the functions of intonation in the literarylanguage.

4. THE ANALYSIS OF THE TEXTUAL STRUCTURE

In the preceding section we analyzed each single stimulus of the experiments(word groups, phrasing units, sentences or whatever combination of those) fromthe point of view of its being embedded in a specific discourse situation of itsown right. We will now proceed to consider the text as a whole and to observe thespecific structure of the rule-governed concatenation of the units to a well-formedtextual whole.

What kind of structure does the narration of Anastasija (henceforth A), aboutthe near-fatal illness of her son have? What is the nature of the plot, how doesit follow or restructure the objective sequence of events? And, more specifically,

PROMINENCE IN A RUSSIAN DIALECT TEXT 273

what gives rise to a discourse situation where alternative accentuation is possiblewhile this freedom is in other cases restricted or excluded? What is the differencein the textual functions of these alternatives and what are the semantic andinformational consequences of these decisions? We will start with a roughdescription of the text and the accentual patterns that have been assigned to it,by the author on the one hand, by our subjects on the other.

We will, at the beginning, consider the acceptability of the basic text inwhatever version it appears. The narrative of A has been accepted by the subjects,speakers of the standard language and not of the dialect spoken by A, both inthe written form and as an oral text. Both versions can be classed, consequently,as coherent text received with acceptance and a full understanding by membersof the Russian language community. Those cases in which the experimentalresults exhibit differences in the preferred accentuation of certain portions of thetext should not automatically be ascribed to differences in the intonational rulesapplied. The different accentual positions can be described, as we will see later,with one and the same set of rules. In any case, we can infer that there is a conflictin terms of accentuation, but not in terms of acceptability.

In the preceding section we have undertaken a first step towards analyzinginstances of divergence between the prominence assignment of A and that ofour subjects. The categories we used in this approach were the following: dialect,pragmatic mode – that is, short interlocutor distance (svoj mode) – and a tendencytowards non-neutral word order and intonation. As opposed to this, there isstandard language, cužoj mode, and neutrality in word order and intonation onthe part of the subjects. This opposition of features does not correspond to twodifferent codes; they do not reflect different or complementary sets of prosodicmeans. We can assume a common basis of intonational categories even in caseswhen the exact shaping of the contours would be different as a consequenceof the variety between dialect on the one hand and standard language on theother.

The three layers of description – svoj/cužoj mode, neutral/non-neutral intona-tion and neutral/non-neutral word order – are a good first approximation for de-scribing the overall situation of convergence and divergence coming to light in ourexperiments. When making a decision a speaker or hearer shows himself to be ad-hering to one of the oppositions and reveals himself as far from or near to one ofthe two opposed discourse types. His decision seems to be relatively unrestricted;but we do not see, for the moment, the textual preconditions that allow him, with-out violation of the textual structure, to place accent on alternative places. Thestructure of text, according to our postulate, remains untouched and acceptable.So the text itself has to exhibit the possibilities for the alternatives in question.When we examine the structure of A’s narration, we can see that it reveals a high

274 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

degree of transparency and a well established order of related events. She relatesthe following:

0001 A near-fatal illness has affected her son; this is the overalltheme of the narrative as a whole.

0002+0005a She reacts to this event with concern.0005b–0008b It turns out to be difficult to find help in the isolated and

deserted village.0012 A female neighbour tries to get help by telephone.0015a+b They call the doctor.0022–0023b Different means of transportation are used to bring the son to

the airport.0024–0029 They meet the doctor and take the son to the hospital.0030–0032a An immediate operation saves the life of the son.By transparency we mean that all events, following a dramatic pattern, are

represented in the text in their logical, causally connected order. Independentof dialect, mode and neutrality in the field of word order and prosody, the plotis perfectly well understandable. The simple outline of the narrative remainsuntouched by the different choices of prominence assignment being made by thetwo parties, the author and the receivers, that is, the group of subjects of ourexperiments.

We stated in our first attempt at analysis that there is a general tendency tobe found in the behaviour of our subjects to observe neutral positions for theirprominence assignment wherever this is possible. By speaking of a tendency weadmit that decisions deviating from this principle are not to be excluded. Theprominence type favoured by A is not unusual or totally avoided by the subjectsof our experiments; it belongs also to the standard means of expression. But thisis, on the other hand, also true for A: She prefers one of the alternatives, but sheknows and makes use of the opposite alternative also. She is familiar with cužojmode and neutrality. This can be shown by examining the beginning of the text,given below in a simplified version.

0001 Zabolel syn.0002 Jazva razlilas’.The subjects agreed in their tendency to choose syn and jazva as the most

prominent words in both versions, which means that they also agreed with thechoice of A. What principles of accentuation can we observe here? The firstsentence is the neutral form of a text beginning; cf. the type звонит телефонin the study by Исаченко (1967). The continuation in the form of 0002 is alsowell-known in the literature; it corresponds with the type Вдруг автомобильостановился – отказал мотор in Mehlig (1994, 184). The first sentence inthe latter example depicts an event that needs explication, and the second sentencefills this gap, the explicative nature of this sentence being signalled by the stress on

PROMINENCE IN A RUSSIAN DIALECT TEXT 275

the subject in front position. This description is valid for the convergent behaviourof A and our subjects; it is a semantic category independent of mode, style, anddialect.

We consider next a typical case of divergence.0030 V bol’nicu privezli.0031 Operaciju sdelali.

In the first sentence, 0030, we find convergent prominence on privezli, while in0031 subjects favoured operaciju but heard A accenting sdelali. Their own choice,it can be noted, does not affect their perception. Given the short interval and theproximity of the two sentences it seems very unlikely that A made a shift in modethat the subjects did not follow. The two models being followed here are by nomeans isolated cases and they can serve as a key to the mechanism of alternativeassignment and the underlying semantic motivation.

To bring light into the nature of this key we will return to the analysisof 0022 given in the preceding section. There we also found an instance ofconvergent assignment, with A accenting the finite verbal form, refusing, on thebasis of her individual subjective decision, to put an accent on the referentialexpressions accented by the non-dialect subjects. As a result, A gives thereferential knowledge, out of the range of prominence, the status of knowledgeavailable to the listener by way of imposition. Adopting the view presented there(cf. 3.2.2) we can infer that sentence 0031 gives the answer to the questionOperaciju delali ili net? By putting this disjunctive question into the matter ofcurrent concern of the listener, A gives him actual possession of the knowledge ofthe whole predication with the exception of only one element, which is the truthvalue of the predication. Presupposed is the predication as such, and the utteranceof 0030 adds only one element to the discourse situation, namely that the predicateis true.

The observed function can be described as follows. The impositional partitionof information is given a grammatically proper expression in the field of intona-tion. This type of sentence semantics is described by Höhle (1992) for Germanand by Mehlig (1994) for Russian, under the heading of verificational type ofinformation structure. A verification is uttered as a sentence with an accent onthe finite verb; its focus is placed on the – positive or negative – truth value ofthe whole predicative structure. Having in hand an adequate grammatical descrip-tion of a well defined semantic category, we can describe cases like the above asinstances of Russian grammar, independent of standard or dialect.

At least one question remains open: What is the motive for a narrator to givethe form described to a sentence in a series of sentences representing a series ofconsequent events? One possible reason is to be seen in the series 0030–0031:the events are not seen in their unique individual consequence, but as following aschematic independently motivated order. Following this order, A doesn’t develop

276 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

the plot by dealing with each single part of it as new information added to thebulk of information accumulated so far. She confirms, instead, that the eventhappened in the way that the schematic order would have predicted it. The plotof the narrative is replaced by a frame (see van Dijk’s 1972 version of this widelyaccepted concept) consisting of a certain number of typical components; that oneof the components has taken place does not have to be related, but has to beconfirmed.

This form of textual organization can be observed, in our text, in the passagedevoted to the transport of the son to the airport (0022–0023). This is a pieceof the narrative where there are mostly divergent assignments (see Table 6). Thesubjects’ preferred places of accent are referential expressions; A prefers to puther accents on the verbal components, confirming in this way the appearance ofevents as predicted by a transportation frame. This alternative is far from beingfollowed rigidly by one of the parties involved in our experiments. They displaydifferent tendencies in terms of presupposing whether information to be relatedhas or has not been placed in the concern of the hearer, making use of impositionand of the canonical rules of sentence semantics.

5. SUMMARY

Alternative ways of assigning prominence within one and the same coherent textare possible and do exist; they do not affect the acceptability, the understandingor the interpretation of its content. Our analysis has shown that the motivationfor the choice is in these cases to be seen, on the one hand, in the choice ofinterlocutor treatment (svoj vs. cužoj). There are, on the other hand, differingways of organizing the informational structure of the text, described in terms ofexplication and verification. The importance of this observation lies in the factthat both alternatives use the rules belonging to one and the same grammar anddo not have to be explained as a switch from one set of prosodic rules to theother.

The two groups of language users, the dialect speaker and author of the originaltext on the one hand, and the speakers of the standard language and subjects ofour experiments on the other, show a clear tendency to favour different solutions.This can be initially explained as a difference in textual structuring. The standardgroup prefers the pieces of text to be concatenated in accordance with the flow ofevents, each piece adding something new to the already established situation. Thedialect narrator, however, sees her role as someone who has collected the events ina pool accessible to both speaker and listener. Her narration is a process of pickingunits out of the pool and confirming that those events indeed happened. To takean example: In the pool this dialect narrator uses there is also the informationthat in her house at that moment was a little girl (see sentence 0007b). At the

PROMINENCE IN A RUSSIAN DIALECT TEXT 277

moment of relating this detail there is no evidence regarding the importance ofthis circumstance for the plot. But it may have been important for the frame ofaccumulated trouble as it adds some extra reason for the mother to be bothered ina situation already troublesome enough.

These observations also have relevance for the problem of dialect intonationin general. Prominence assignment is just a part of the intonational system as awhole; it is responsible for the information contained in the respective intonationunit and its relation to the knowledge system of the interlocutor, be it objectivelygiven or imposed by the speaker. In the process of structuring the discourse itis evident that the rules of prosody as well as the other rules of grammar areexploited in different ways according to the intention of the speaker. The piece oftext we made use of exhibits some alternative ways of exploitation that do not alterthe written representation of the text. This is certainly an extraordinary case. Andit gave us the possibility of checking the hypothesis that these alternatives can beexplained within one and the same grammar. They maintain their communicativefunction, even when the interlocutors speak different varieties or dialects of onecommon language.

Differences in prominence assignment have been shown to be tendencies, notoppositions of mutually exclusive intonation systems. For the study of dialectintonation this does not in general exclude the possibility that such oppositionsmight exist. But when confronted with differences in intonational behaviour weshould bear in mind the approach we favoured for our analysis: that there might bedifferences in the structuring of discourse, different modes of positioning eventsin relation to one another, differences in establishing the position of an event inthe text, but also differences of assessing the position the knowledge concerningan event has in the hearer’s mind. These differences might be individual, but theycan also be specific for a social or cultural group.

“Accent is predictable – if you are a mind reader”. This motto of Bolinger hasmotivated us in our search for these differences, not in the realm of grammaralone, but also in our efforts to take into account speakers’ intentions. From ageneralized model of discourse we may add: Linguists must not only try to readthe speaker’s mind. They also have to reconstruct the way the speaker is readinghis interlocutor’s mind. Does he presuppose other knowledge sets in the hearers’mind because of his living in a different context, the sphere of his rural world?Or does he make use of a different type of storytelling based on a different,orally oriented cultural background? The domain of the information structure ofdiscourse can help us to clarify those questions, and in this domain we expect newand stimulating insights into the realm of dialect speakers’ use of intonationalmeans.

278 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

NOTES

i The authors acknowledge the invaluable assistance of Tatjana Šerstinova from St. Petersburg StateUniversity who performed the experiments with expert skill.ii Note that the number of words the subjects were allowed to mark as prominent was not limited toone per token. Hence the total number of perceptually-based markings in the second sentence in 0007aequals 28 (=18 + 10), despite the total number of subjects being 25. Evidently, at least three judgedboth the predicate (4) and the adverbial phrase (5) to be accentually prominent. In the same sentence,when the subjects were asked to mark the words they expected to be prominent, their markings totalled25 (=6 + 19), thus indicating a complementary distribution of their responses. prominent word in0005b, or that the clitic ved’ is accentually prominent (in 0022a). Such puzzling judgments may simplyhave been “slips of the pen”.iii No attempt was made to identify the phonetic correlates of accentual prominence as understoodby the auditors, especially in the written experiments. They could have been reacting to amplitude,Fo, intonational contours, syllable length, or any combination of these and other features. Some ofresponses were confounding: it is difficult to imagine, for example, what would lead one to assume thatthe predicate (3) would be (the most?) prominent word in 0005b, or that the clitic ved’ is accentuallyprominent (in 0022a). Such puzzling judgments may simply have been “slips of the pen”.iv In the text presented to the subjects in written form, the particle was given as -to. Repeatedlistening, together with the spectrogram, actually revealed tu in the rendition of the narrator. It remainsunclear, given this segmental composition, if tu is actually cliticized to (5), or if it constitutes thebeginning of the next utterance. This ambiguity, however, does not affect our analysis.v This verb form was in fact rendered by the dialectal narrator as pomr’o; the morphology in thewritten experiment was normalized.vi Nearly half, consequently, marked both words, with a slight preference for Valerko (da ved’).vii The clitic da is incorporated into the next word on the basis of th same considerations as thosediscussed in connection with token 0012b.viii The clitic už, which the subjects unanimously evaluated as unaccented, is attached here to thepreceding fully accented word tam on the basis of the same considerations as those discussed inconnection with token 0012b.ix For a formal description of narrative intonation in the cužoj mode see Yokoyama (2001). Therelevant contour here is that of the last syntagm.x In the Transactional Discourse Model such referential knowledge is placed by the speaker into theintersection Ca∩Cb, and as such it cannot be assigned the HL contour tone that concludes the coreintonational formula for the intonation of the distant cužoj mode.xi The relative prominence of Valerko and Griška vis-à-vis their respective predicates is noticeablewhen the actual percentage, rather than the simplified numbers, is considered: Valerko = 80% versuspomrë(t) = 72% (both 4), and Griška = 56% versus poezžaj = 44% (both 5).xii The exact intonological formula in question will not concern us here, as it would be possible toarrive at it only after a general description of the intonational system of Russian in general and of thisdialect is in place.xiii The difference between the markings in the experiments with written and spoken material arenot significant (n.s.), if p > 0,05; significant (s.), if n.s. p ≤ 0,05; highly significant (h.s.), if p ≤ 0,01;very highly significant (v.h.s.), if p ≤ 0,001.xiv For this fragment, we got the reactions of only 24 subjects.

PROMINENCE IN A RUSSIAN DIALECT TEXT 279

APPENDIX I:

Table 5. The fragments of the text in Kasatkina (1991, 51–53), that were used for the experiment (inthe form they were – without the file name in the first column – presented to the subjects)

280 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

APPENDIX II:

Table 6. The assignments of the most prominent words: the results of the experiments with writtenand spoken material (for comfortable reading extraposed from the main text)

PROMINENCE IN A RUSSIAN DIALECT TEXT 281

282 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

PROMINENCE IN A RUSSIAN DIALECT TEXT 283

APPENDIX III:

The graphs depicting the energy (upper profile) and the Fo (lower profile) offragment 0012

284 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

PROMINENCE IN A RUSSIAN DIALECT TEXT 285

BIBLIOGRAPHY

Bolinger, D.: 1972, ‘Accent is predictable (if you’re a mind reader)’, Language 48, 633–644.van Dijk, T. A.: 1972, Some Aspects of Text Grammars, The Hague.Elsner, A.: 1999, ‘Prediction and perception of focal accents’, Proceedings of the 12th ICPhS, San

Francisco, 1549–1552.Fant, G., Kruckenberg, A.: 1999, ‘Prominence Correlates in Swedish Prosody’, Proceedings of the

XIIth ICPhS, San Francisco, 1749–1752.Höhle, T. N.: 1992, ‘Über Verum-Fokus im Deutschen’, Jacobs, J. (ed.), Informationsstruktur und

Grammatik, Opladen, 112–141.Knipschild, M., Sappok, Ch.: 1991, ‘Akustische Zeichenverarbeitung durch Sona und Versteu’,

Fortschritte der Akustik – DAGA, Bad Honnef, 1045–1048.Lohse, H., Ludwig, R., Röhr, M.: 1982, Statistische Verfahren, Berlin.Mehlig, H.-R.: 1994, ‘Explikative Äußerungen. Überlegungen zur Informationsstruktur’, Mehlig,

H.-R. (ed.), Slavistische Linguistik, München, 181–224.Molnár, V.: 1993, ‘Zur Pragmatik und Grammatik des Topik-Begriffes’, Reis, M. (ed.), Wortstellung

und Informationsstruktur, Tübingen 155–202.Schütze, F.: 1983, Biographieforschung und narratives Interview’, Neue Praxis 13, 283–293.Streefkerk, B. M., Pols, L. C. W., Bosch, L.: 1999, ‘Towards finding optimal features of perceived

prominence’, Proceedings of the 12th ICPhS, San Francisco, 1769–1772.Yokoyama, O. T.: 1986, Discourse and Word Order, Amsterdam.Yokoyama, O. T.: 2001, ‘Neutral and non-neutral intonation in Russian: A reinterpretation of the IK

system’, Die Welt der Slaven 46, 1–26.Брызгунова, Е. А.: 1977, ‘Анализ русской диалектной интонации’, Высоцкий, С. С.

(изд.), Экспериментально-фонетические исследования в области русской диалек-тологии, Москва, 231–261.

Брызгунова, Е. А.: 1980, ‘Интонация’, Шведова, Н. Ю. (изд.), Русская грамматика, т. I,Москва, 96–122.

Исаченко, А. В.: 1967, ‘Фразовое ударение и порядок слов’, To Honor Roman Jakobson, TheHague–Paris, 967–978.

Касаткин, Л. Л.: 1994, Русская диалектология, Москва.Касаткина, Р. Ф. (изд.): 1991, Русские народные говоры. Звучащая хрестоматия,

часть 1: Севернорусские говоры, Москва–Бохум.Кодзасов, С. В.: 1996, Просодический строй русского языка, Москва.Люблинская, В. В., Саппок, К.: 2002, ‘Восприятие мелодических характеристик ре-

чевых фрагментов диалектной русской речи’, Вольская, Н. Б., Светозарова,Н. Д. (изд.), Проблемы и методы экспериментально-фонетических исследований,Санкт-Петербург, 191–198.

Пауфошима, Р. Ф.: 1983, Фонетика слова и фразы в севернорусских говорах, Москва.Пауфошима, Р. Ф.: 1989, ‘Об использовании регистровых различий в русской фразо-

вой интонации (на материале русского литературного языка и севернорусских го-воров)’, Николаева, Т. М. (изд.), Славянское и балканское языкознание, Москва,53–62.

Пешковский, А. М.: 1928, ‘Интонация и грамматика’, Известия по русскому языку исловесности, Москва, 458–476.

Светозарова, Н. Д.: 1988а, ‘Особенности фразового и синтагматического членениярусской спонтанной речи’, Светозарова, Н. Д. (изд.), Фонетика спонтанной речи,Ленинград, 144–150.

Светозарова, Н. Д.: 1988б, ‘Степень выделенности отдельных слов и акцентный контурфразы’, Светозарова, Н. Д. (изд.), Фонетика спонтанной речи, Ленинград, 161–165.

286 M. KRAUSE, C. SAPPOK, O. YOKOYAMA

Скорикова, Т. П.: 1997, Русская научная речь. Звучащая хрестоматия, Санкт-Петербург–Бохум.

Ruhr-Universität Bochum, [email protected]

UCLA; Department of Slavic Languages & Literatures, Los [email protected]