Towards a codification of variation in Russian word stress

1. Introduction

In this article a new approach towards a more complete codification of word stress variation in Russian is offered, the goal of which is a descriptive and predictive dictionary providing both synchronic variation and directional tendencies for such words, as well as other necessary information (examples of such a dictionary are given in a table in the conclusion). By their very nature, stress variants represent some degree of ambivalence, which, presumably, in the future will be resolved in one way or another. This much has already been shown by the present author in looking at the development of suffixal stress from the eighteenth to twentieth centuries, so that clear evidence of a general shift from a morphemic stress model to a ‘rhyming’, uniform syllable model has been observed (see, for example, Lagerberg 1999).

Variation in Russian stress is a subject which has to date received insufficient attention, or, at least, attention incommensurate with its level of complexity and importance for the language. Variants, in general, are defined by Gorbachevich (1978, 17) as ‘регулярно воспроизводимые видоизменения одного и того же слова, сохраняющие тождество морфолого-словообразовательной структуры, лексического и грамматического значения и различающиеся либо с фонетической стороны (произношением звуков, составом фонем, местом ударения или комбинацией этих признаков), либо формообразовательными аффиксами (суффиксами, флексиями)’.¹ More precisely, words with variation in stress allow the possibility of two, or occasionally more, syllables on which the stress may fall in a given word form without any difference in meaning (though stylistic differentiation may, of course, be possible). Thus, for example, the plural short form of the adjective ве́рный ‘true, faithful’ is given in standard lexicographical sources (e.g. Zaliznjak 1977 and ORD 1993) as either ве́рны or верны́.

From the outset we should underline that, in essence, we are dealing with two distinct areas of stress variation in Russian as a whole. Firstly, variation may occur in one or more inflected forms of a word, as in the latter example (cf. the other short forms of ве́рный with non-variable stress, viz ве́рен, ве́рно, верна́), or, for example, in the case of the word вода́ ‘water’ the oblique plural cases are given (e.g. Zaliznjak 1977) as either вода́м, вода́ми, вода́х or во́дам, во́дами, во́дах (the latter forms, however, marked as устар. in the said source), but only вода́ for the nom. sg., во́ду for the acc. sg. etc. Secondly, in contrast, a word might be characterised by stress variation in all its forms: essentially this results from variation occurring in the base form (i.e. nom. sg. of nouns, nom. sg. masc. of adjectives, the infinitive of verbs). Thus, if we take the noun творог ‘curd cheese’, which in most contemporary sources (e.g. ORD) is given as either тво́рог or творо́г, one finds that one is dealing with two different accentual paradigms: in the case of тво́рог it is paradigm a (fixed stem stress) and in the case of творо́г it is paradigm b (fixed ending stress), to use the commonly applied system.² Thus, the former base form results in all inflected forms having fixed stem stress on the initial syllable (e.g. gen. sg. тво́рога), and the latter in all forms having ending stress (e.g. gen. sg. творога́).

A subtype of the latter type of stress variation occurs when stress falls on different syllables of the stem (but not the ending), in which case one is dealing with one stress paradigm (a or fixed stem stress), but with the stress falling on a different stem syllable in all forms of the paradigm, e.g. мы́шление/мышле́ние ‘thinking’, gen. sg. мы́шления/мышле́ния. This type of stress variation plays a prominent role in word-formation where words are derived by certain suffixes: typically, words containing the same suffix may either form uniform categories in terms of stress (so called ‘dominant’ suffixes such as -ист with stress always, that is, without any exception, on the same, in this case, final stem syllable, e.g. коммуни́ст ‘communist’, шахмати́ст ‘chess player’), or a more complex ‘pattern’ may occur with stress occurring in two or more different syllabic positions in different words. An example of this is provided by the verbal suffix -ировать which can be stressed on the final or pre-penultimate syllable in different verbs: thus маскирова́ть ‘to mask’ versus телефони́ровать ‘to telephone’, and there are even individual verbs with both stress positions permitted in standard sources (without any distinction in meaning, but with possible different stylistic evaluation), e.g. татуирова́ть/татуи́ровать ‘to tattoo’ (Zaliznjak 1977). Such, therefore, are the main types of stress variation encountered in standard Russian.³

2. Sources of stress variation

The starting point for any dictionary of stress variation is, of course, the data itself. There are two main sources of such data, surveys of native speakers and dictionaries of stress and pronunciation. While the former offer, in theory, objective and current usage, the difficulties involved in compiling sufficiently broad (in terms of both the number of respondents and the lexemes or forms whose stress is investigated) surveys are considerable, and currently there is insufficient data of this kind on which to base a wide-ranging analysis of stress variation in Russian. Surveys of Russian stress have had considerable variation in size, with only one (that of the Русский язык и советское общество research group of the Russian Language Institute reported in Krysin 1974) reaching a size of any significantly large proportions. The survey sizes to date known to the present author are: 3 (Agrell 1917), 150 (Vorontsova 1959) (though not every question receives a total of 150 responses), 225 (reported in Kolesov 1967), 75 (reported in Pirogova 1967), approximately 250 (reported in Gorbachevich 1978 without an exact number of informants given), 5000 of which 4300 were processed in morphology/stress and the largest survey to date (reported in Krysin 1974), 4 (Strom 1988), 143 in Larsson (2006), 21 (reported in Ukiah 2001 and elsewhere), 15 (Lagerberg 2005), 20 (Lagerberg 2008a, 2008b).

Lexicographical sources (e.g. Zaliznjak 1977, Esʹkova 1994 and Gorbachevich 2000), though they may lag behind surveys by some years in terms of the currency of the data itself, nevertheless offer a more substantial corpus of data from which data on stress variation can be extracted. The issue of norm is important here. As Lehfeldt (2006, 110 ff) points out, the norms of standard Russian usage have been relaxed to some extent during the last twenty years or so, allowing for certain categories of stress usage to be mentioned without overt pejorative evaluation. In Esʹkova’s dictionary (Esʹkova 1994), for example, five categories of variation are included on a scale ranging from, at the most tolerant level, a) absolutely equivalent variants (e.g. и́скристый vs. искри́стый ‘sparkling’), b) two normative forms of which one is basic and the other allowed or allowed/obsolescent, to the non-normative c) not recommended or not recommended/obsolescent, then d) incorrect, and, finally, e) the rather colourfully named ‘грубо неправильно’, e.g. мага́зин vs. correct магази́н ‘shop’. Esʹkova (1994) also includes what are generally considered dialect forms, for instance, by Gorbachevich (1978), under these latter two categories.

In the conclusion we provide examples of how a dictionary of stress variation might appear. Using one of the latest and most extensive of these above sources (Gorbachevich 2000), descriptions of the stress variation in question, as well as any normative evaluation relating to it, are provided. In particular, the assessments неправильно, не рекомендуется and допустимо are taken respectively as evidence of non-normative usage in ascending quantitative order, that is to say that such usage is taken to be evidence of variance and its varying potential to co-exist with, influence and even replace standard usage in the future.

3. A typology of stress variation

Before a more systematic analysis of stress variation can be conducted, it is expedient to delineate different types of variation which exist in the total corpus of Russian words, since this will enable us to determine more accurately relevant tendencies and probable future directions in stress change.

In a previous article (Lagerberg 2007) the following main categories of stress variation were identified:

• Inflectional variation: alternative stress in at least one form in the paradigms of nouns, pronouns, numerals, adjectives, adverbs and verbs (including participles and gerunds).
• Derivational variation: in contrast to inflectional stress, this type of variation a) occurs, of course, in derived (mainly suffixed or prefixed forms) and b) generally occurs not in particular inflected forms, but in the initial form (nom. sg. of nouns and (masculine long form) adjectives, infinitive of verbs), giving rise to variant inflected forms throughout paradigms, i.e. in every inflected form of more than one syllable (where they exist).
• Free variation: a relatively small number of cases of stress variation where a non-derived word, or, at least, a word whose morphology is perceived to be non-derived, has two free, legitimate stress positions in its initial form (and, therefore, two different stress paradigms) with no difference in meaning and, therefore, no apparent motivation, e.g. тво́ро́г ‘curd cheese’.
• Semantic variation: cases of stress variation when a different stress is connected to a different meaning, i.e. these are homographs, e.g. а́тлас ‘atlas’ and атла́с ‘satin’.
• Dialectal variation: an alternative stress position identifiable as non-standard and regional.
• Professional stress: identified by Zaliznjak (1985, passim) as a variant stress (which, he states, is often ‘non-trivial’ stress, i.e. desinential stress or mobile stress) in the spoken (specialised) vocabulary of certain professions; in fact, a kind of phonological jargon, e.g. pattern-b stress in масса́ж ‘massage’ amongst masseurs, as opposed to stem stress in the standard language (68–69).

Of these six basic types of stress variation two are inherently close to each other, namely inflectional and derivational, since both essentially represent intermediate stages in an ongoing analogical process. Though inflectional stress variation is generally more complex and volatile than derivational, both types appear always to have an ultimate goal towards which they are proceeding, namely uniform stress among all members of the given sets, and to that extent both types represent unstable, ambivalent stress. The modus operandi is different in each, and the sets too are, of course, different; for inflected words the set in question is determined by the stress type of the word. Thus, if it is a pattern f word (e.g. губа́ ‘lip’) it is immediately in tension with a general pull towards a binary singular–plural opposition in words with mobile stress, since it maintains mobile stress in the plural sub-paradigm (nom. pl. гу́бы, dat. pl. губа́м). For suffixed words, the set is determined by the suffix in question. Thus, if, for example, a word contains the suffix -ировать and has final stress (e.g. премирова́ть ‘to give a bonus’), then it is already in tension with the general tendency of that suffix (which can generally be established empirically by diachronic analyses) towards stress on the pre-penultimate syllable (e.g. variant stress преми́ровать). While free variation and semantic variation, albeit on a smaller scale, share some characteristics with inflectional and derivational stress variation, such as ambivalence and a potential motivation by certain factors towards a single stress position, dialectal and professional stress represent more isolated and generally insulated types of variation which are less likely to have any major effect on the standard language. For words with free variation it is difficult to generalise about possible directional trends other than to suggest factors which operate on the language at the general level. In particular, one can mention Gorbachevich’s (1978, 63) theory of rhythmicity in syntagmas with a typical two–three syllable spacing from one stress to the next, sometimes extending to four syllables – what he calls the ‘critical interval’. In this way, according to Gorbachevich, the freedom of Russian stress becomes something more relative, tending towards the middle/middle-right of any given word unit of three or more syllables, and even, in some cases, leading directly to a shift or variation of stress. Within semantic variation there is the potential for one stress variant to become more dominant (in both meanings) in line with its own higher frequency if there is significantly low or lower frequency in the other variant. Combinations of some of these types are also possible (see Table 1 for examples).

Within the important category of inflectional variation, the following areas of accentual ambivalence can be highlighted:

4. The pragmatic factor and frequency

Two other factors which need to be taken into account in stress variation are the so called ‘pragmatic factor’ and that of frequency. The former is defined by Zaliznjak (1985, passim) as the effect on stress in underived and derived words of their relative assimilation or lack of (respectively освоенность/неосвоенность) by native speakers. Zaliznjak states (1985, 12): Для взрослого носителя языка освоенными являются слова повседневной обиходной лексики, названия предметов, связанных с его профессиональной деятельностью, слова, специфические для его микроколлектива или ограниченной среды. Неосвоенными являются, в частности, названия предметов из жизни других стран или прошлых эпох, термины незнакомых данному человеку профессий, слова чужих микроколлективов, для большинства носителей языка – также все слова возвышенно-книжного и архаизирующего стиля. Неосвоенным для данного человека является также всякое слово, которое он только что узнал; однако через некоторое время, иногда довольно короткое, он может вполне освоить это слово. (For an adult native speaker assimilated words are words of everyday colloquial vocabulary, names of objects connected with his or her professional work, words specific to his or her micro-group or confined milieu. Unassimilated words are, in particular, names of objects from the life of other countries or previous eras, terms from professions which are unfamiliar to the person in question, words of other micro-groups, and also, for the majority of native speakers, all words of an elevated bookish and archaizing style. An unassimilated word for the person in question is also any word which he or she has only just become familiar with; however, after a certain period of time, sometimes quite short, he or she can assimilate this word fully.) Assimilation is, of course, not an absolute quality, it is dependent on each individual and is variable for that individual during his or her lifetime and is, thus, an extremely difficult thing to measure accurately.

The effect of the pragmatic factor, according to Zaliznjak, can be divided into two main categories: underived and derived words. For the former, an unassimilated word displays a tendency towards ‘trivial’ stress (fixed stem stress), while for assimilated words ‘non-trivial’ stress (desinential/mobile stress) is the general rule. The logic of this is reasonable and follows the argument to be made for frequency in underived words (assimilation being closely related to frequency, of course, since low frequency words are, or will become, unassimilated for most speakers, and vice versa), though one might take exception with pattern b feminine nouns, since in many cases they are not of the assimilated type. Indeed, feminine words with fixed ending stress (pattern b) are confirmed as a low frequency group by Cubberley (Cubberley 1987, 34–35). Cubberley discovers that only 5 of the 357 most common, according to Zasorina (1977), feminine nouns in Russian have pattern b, even though they represent overall in Russian the second most common group (after the pattern a fixed stress pattern) according to Fedjanina (1982). For derived words the argumentation is, in our view, far more spurious. Essentially Zaliznjak (1985, 79) takes the view that assimilated words yield stress to the suffix, while unassimilated words are conservative and retain original or older stress positions. For example, the adjective мускулистый ‘muscular’, in Zaliznjak’s view, has a more colloquial suffixal stress (мускули́стый) by virtue of the fact that it is assimilated; the conservative, morphemically inherited stress му́скулистый represents the literary norm. This would appear to fly in the face of logic, since assimilated words, unless they have fallen into disuse and become again assimilated at a later date, are by their nature conservative in their stress. Because assimilated words are in constant use, their stress is constantly reinforced, even though it may go against the general trend for, for example, a given suffix such as -истый (which tends towards the stress -и́стый). In the same way, high frequency/assimilated underived words are able to maintain complex mobile patterns in the consciousness of most speakers purely through their constant use, while those which become less frequently used and less assimilated gradually adopt more simple (fixed or columnar) stress patterns. For this reason, I view sceptically Zaliznjak’s pragmatic factor vis-à-vis derived words, while considering it valid for underived words.

Closely connected to the pragmatic factor is another important factor which needs to be taken into account in stress variation, namely frequency, though here there are also considerable difficulties in doing so, and by no means all scholars are convinced that it has any discernible effect. For example, Marklund Sharapova (2000, 36–37) writes: ‘It is doubtful what role frequency plays in stress variation and change, and the conclusions … are somewhat contradictory.’ In contrast, though not dealing with stress variation, but phonological variance, File-Muriel (2010, 19) finds that ‘lexical frequency plays a clear role in sound reduction and that these rates are graded across the lexical frequency scale’. Tornow (1984, 469) states that the more frequent a word, the more likely it is to have stress variation: ‘Akzentdubletten sind bei den häufigsten Wörter fünfmal so gebräuclich wie bei allen Wörtern. Je öfter ein russisches Lexem verwendet wird, desto eher in akzentuell verschiedenen Formen. Der scheinbare Zusammenhang mit der Mobilität erklärt sich durch die Häufigkeit der Mobilität.’⁴ In simple terms the argument is that, as Tornow states it, higher frequency in inflectional stress is connected with a higher probability of more complex stress patterns (i.e. mobile patterns) as well as a concomitant higher probability of stress variants (one of which may be innovative and/or deprecated) occurring in individual inflected forms (generally not the base form). Higher frequency in suffixed words, by contrast, in the view of the present author, is connected with ‘anomalous’ stress, i.e. stress which differs from the general model for a given suffix (i.e. the reverse of the pragmatic factor), while decreasing frequency, on the other hand, appears to increase the probability of uniform or ‘analogous’ stress; between the two, however, there appears to be an intermediate area where variants occur, basically lexemes which are currently ambivalent in terms of their stress, though ultimately heading towards uniform stress.

An example of the latter type of variation can be seen with the combining form -лог. Although for animate nouns with this suffix, penultimate stress has been generalised (e.g. гео́лог ‘geologist’), for non-animate nouns there is a certain tendency towards final stress (e.g. моноло́г ‘monologue’). However, a survey conducted by the present author (Lagerberg 2008b) revealed a greater amount of variation than is reflected in normative sources. Five out of eight words surveyed displayed varying degrees of stress variation, but all basically tending more towards a preference for penultimate, pre-suffixal stress (e.g. аполо́г to апо́лог ‘apologue’). Only three words received uniform responses for stress, namely диало́г ‘dialogue’, моноло́г ‘monologue’ and проло́г ‘prologue’, all with a relatively high frequency count.⁵ Thus, the survey showed quite clearly that higher frequency in this case appears to play a role in preserving the older stress position. On the other hand, it does not operate as an absolute guarantee of ‘stress preservation’, but rather a contributing factor: каталог ‘catalogue’, a word with a relatively high frequency, received only slightly more than half the responses in the survey as катало́г, the other informants favouring ката́лог.

It is, therefore, important to note that frequency is not an absolute test of stress position, nor a guaranteed factor to fall back on in difficult cases of stress – there certainly are other factors which could play a role, such as analogy, semantics and the influence of other related words. However, it seems clear that to ignore the frequency factor is also to restrict oneself to an incomplete understanding of what is currently taking place in the spoken Russian language. There is certainly a higher probability of ‘anomalous’ stress being retained in such cases where the word has a significant level of frequency. Where exactly that level begins is still a moot point which probably can never be gauged exactly, but it remains, nevertheless, a factor to be considered. On this point, File-Muriel indeed rejects the proposal of absolute frequency ranges (viz high, mid, low), though, indeed he uses the notion of high frequency throughout his argument (File-Muriel 2010, passim). He finds that lexical frequency ‘plays a clear role in sound reduction and that these rates are graded across the lexical frequency scale’ (19). Bybee (2006, 715) notes that it is impossible at the moment to specify ranges for extreme high, medium, and low frequencies. She suggests, however, that as more empirical studies appear, absolute frequency ranges for each phenomenon will eventually be specifiable. I suggest that if lexical frequency is considered as a scalar variable, this will become a non-issue, given that frequency in scalar terms produces a valid number for each word or morpheme relative to other words/morphemes in a given corpus. Formalizing ranges of frequency ranges misses the important point that frequency is a relative variable which exercises gradual effects on cognitive representation. Considering lexical frequency as a scalar variable allows one to make empirical claims regarding the gradual effects of lexical frequency, which is not possible following a two-way classification (i.e. high frequency versus low frequency).’

5. Conclusion

In this paper an attempt to codify variation in Russian word stress has been made: a typology of variation is presented, as well as some of the most important areas of variance within inflectional stress, the most complex of these types. The so- called ‘pragmatic factor’ and the effect of frequency, somewhat nebulous phonological areas, are also included in this discussion as necessary tools in a more complete analysis of the dynamics of stress variation. The ultimate goal of this approach is the production of a dictionary of Russian stress variation which will give both full, synchronic descriptions of the nature and the type of instances of variation, as well as information on the possible directions of any future change and the potential effect on that change by the pragamatic and frequency factors, and to this end an experimental dictionary of stress variation, which uses selected lexemes from Gorbachevich (2000), is presented. In the first column the word in question is listed, indicating its stress when it contains more than one syllable and has no stress variation in this initial form. In the second column the details of the stress variance, as described in Gorbachevich (2000), are given. Owing to the nature of the task, details here are kept to a minimum, but enough information is given to deduce further variation, i.e. in other inflected forms. In the third column the type of variation is given. In the fourth column the direction in which the variance can be concluded to be proceeding is described. The assessments неправильно, не рекомендуется and допустимо, rather than simply being labelled and dismissed as deviations from the norm, are taken respectively as evidence of non-normative usage in ascending quantitative order. In the fifth column we indicate the potential for frequency and the pragmatic factor to be playing a role in the latter (assumed) directional change.