Abstract
Using written texts elicited from students with different proficiency levels, we studied the acquisition of nominal cases in Russian as a second language. We established the order in which cases were acquired (nominative, locative, accusative, genitive, instrumental, and dative), as well as certain characteristics of their acquisition trajectories. We argue that the order of acquisition is determined by two groups of factors: how essential a given case is for successful language use and how complex it is, both semantically and morphologically. Many important differences between first and second language acquisition were identified. The maturation of the case system was observed both in the number of correct forms participants produced and in error rates. The distribution of correct forms changed, gradually approaching that found for native speakers. With growth in proficiency, the incidence of different error types declined, and the relative frequency of errors more characteristic of first language acquisition increased.
I Introduction
In this article, we aim to trace the development of the nominal case system during the acquisition of Russian as a second language (L2). This system is very complex: Russian nouns are inflected for six cases and two numbers and have different sets of inflections depending on their gender, animacy, and stem type. Although there are many studies on first language (L1) acquisition of Russian cases, relatively little is known about L2 acquisition.
As we show below, L2 research has focused either on the acquisition of particular cases (Gor et al., 2017, 2019; Kempe and MacWhinney, 1998, 1999), or on a single stage of case-system development (Rubinstein, 1995a, 1995b). We studied the development of the whole system, including the order of acquisition of all six cases, from beginner to advanced levels. This has not been previously done either for Russian, or for any other morphologically rich language.
The following section reviews the central features of the Russian case system that must be acquired by L2 learners in order to use cases correctly. Then L1 and L2 studies on the acquisition of Russian cases are presented. Finally, we turn to our study.
1 Russian nominal case system
This section contains a brief description of the Russian nominal inflectional morphology. Standard Russian has six cases: nominative, genitive, dative, accusative, instrumental, and locative (sometimes also called prepositional) and two numbers: singular and plural. Nouns may have different inflections depending on their gender (masculine, feminine or neuter), stem type, and animacy. Based on their set of inflections, nouns are divided into several inflectional classes or declensions.
Although declensions are present in every theoretical description of Russian inflectional morphology and in schoolbooks for native speakers, they are not generally used in teaching Russian as an L2. 1 L2 learners acquire case inflections only in relation to gender, as it is shown in Tables 1 and 2 (after Kempe and MacWhinney, 1998). This system is similar to the system with four declensions, but ignores a small group of masculine nouns ending in -a/ja in nominative singular: they are presented as exceptions.
Case inflections of Russian nouns: Singular paradigms.
Case inflections of Russian nouns: Plural paradigms.
As Tables 1 and 2 make clear, the animacy of the noun determines its ending in accusative case. The choice of inflections (-y/i, -a/ja, etc.) also depends on the final consonant of the stem: velar, affricate, other non-palatalized and palatalized consonants. The symbol ø means no overt inflection: the bare stem ending in a consonant is used. The symbol ’ stands for the letter ‘soft sign’, which usually indicates that the preceding consonant is palatalized, but has a number of other functions.
The Russian case system includes diverse patterns of morphological syncretism. This creates additional obstacles for L2 learners, both when they learn the paradigms and when they try to determine the case and hence the syntactic role of the noun based on its inflection. However, the choice of the right case is even more difficult than the choice of an inflection, depending on the case. The case of a noun may be determined by its role in the sentence and/or by a verb or a preposition that governs it. Some prepositions may require different cases depending on their functions.
Case meanings or functions have long been a matter of debate in the theoretical literature. For example, Jakobson (1936/1984) assumed that all specific meanings of a particular case are closely connected and can be summarized in one invariant. However, the invariants he deduced were very abstract and turned out to be controversial, so many linguists gave up the idea. Thus, Wierzbicka (1980) claimed that multiple meanings of one case have a common core, but there may be no invariant meaning.
The system of case functions used in teaching Russian as an L2 does not rely on these theoretical approaches; it does not aim to reveal the common core of case meanings, only to list the most frequent contexts for every case. Each case is assumed to have main functions, 2 such as:
Nominative: subject of a sentence;
Genitive: possession, negation, quantity, measure, place;
Dative: indirect object, place, direction;
Accusative: direct object, direction, measure, quantity, time;
Instrumental: instrument, means of transport, time, place, manner;
Locative: place, object of speech and thoughts, means of transport, time.
However, the number of functions introduced may be higher, for example, up to 23 for genitive.
In some instances, it is difficult to explain why similar functions are served by different cases. For example, the function ‘means of transport’ may be conveyed both by the locative and the instrumental case, as in (1a–b). As a result, in addition to general rules, L2 learners have to remember cases associated with particular verbs and prepositions (sometimes depending on their functions).
(1) a. exatʹ na poezde goINF by trainLOC.SG ‘go by train’ b. exatʹ poezdom goINF trainINS.SG ‘go by train’
L2 learners start the acquisition of cases with nominative. Other cases are presented in the following order in most Russian coursebooks (e.g. Chernyshov and Chernyshova, 2019; Nakhabina et al., 2015): locative > accusative > genitive > dative > instrumental. A few authors prefer to introduce accusative before locative; in the coursebook used by the participants in the present study (Gorbatkina et al., 2003, 2004), instrumental is introduced before dative.
Relative frequency of different cases may be another factor affecting their acquisition. The counts in Table 3 are taken from Slioussar and Samojlova (2015), whose study was based on the Russian National Corpus (www.ruscorpora.ru). As Table 3 makes clear, the picture is very different for animate and inanimate nouns. In the corpus, inanimate nouns are about three times more frequent than animate ones, but certain animate nouns are very frequent in any native or L2 speaker’s input. Other frequency counts based on different corpus samples (not taking animacy into account) can be found in Kopotev (2008), but the order of oblique cases remains the same: genitive > accusative > locative > instrumental > dative. Only in Josselson (1953) is the accusative more frequent than the genitive.
Relative frequency of cases in Russian (percentages).
Summarizing this section, we can conclude that it is not evident which cases should be easier for L2 learners to master. For example, accusative and genitive are the two most frequent oblique cases, but they are both semantically and morphologically complex. Genitive has the largest number of functions and the greatest variability of inflections (if plural paradigms are considered). In accusative, the choice of inflections is determined by the animacy of the noun, and, except for its main function as a direct object, its functions are far from simple: measure, quantity, direction, time interval. Locative is much less frequent than genitive and accusative, but in most coursebooks it is introduced before them, and it is relatively simple semantically and morphologically. By finding out how often L2 learners try to produce these and other case forms, and how many errors they make at different proficiency levels, we can figure out how these factors affect L2 acquisition.
2 L1 acquisition of cases in Russian
One of the key questions in L2 studies is analysing how – and why – L2 acquisition differs from L1 acquisition. Therefore, in the present section we briefly review existing studies on case acquisition by Russian children. The very first study was conducted by Gvozdev (1948/1961) and was based on the observations of a single child, his son Zhenia. The data collected by Gvozdev are also discussed in Elkonin (1958, 1958/1973), Slobin (1966), and elsewhere. Although the dataset was small, many generalizations were later confirmed in larger studies. Zhenia could distinguish different cases quite early, at the age of two, and Gvozdev observed almost no case mixing. However, Gvozdev noticed mixing of endings within one case (Zhenia used inflections from a wrong declension). Therefore, inflectional paradigms were acquired later than case functions.
Children start with nominative forms, using them in single-word utterances. But as soon as multi-word sentences appear, creating syntactic environments for case distinctions, nominative is used for subjects and accusative for objects. In early studies, there was a debate about which case appears after nominative and accusative: analyses of different datasets suggested genitive or dative. However, after the work of Lepskaya (1988), it was decided that there is no particular order of case acquisition by native Russian children. Lepskaya demonstrated that after the distinction between accusative and nominative is in place, other cases appear very fast and almost simultaneously, i.e. the case paradigm is acquired as a whole (for an overview, see Voeikova, 2011; Voeikova and Gagarina, 2002).
Still, there are remarkable frequency differences between different case forms. Babyonyshev’s (1993) study was based on data from two monolingual Russian children. They were 2.1 and 1.6 years old at the beginning of the study and were regularly recorded during a 6-month period. She counted different case forms in her data. The absolute majority were nominative, and a number of accusative, genitive, and dative forms were recorded. There were very few locative forms, and no instrumental forms were observed. This pattern is consistent with the observations made by other authors.
Notably, children did not use nominative in contexts in which other cases were required; its frequency was due to the prevalence of one-word and two-word utterances at the early stages of language development. In general, the number of case errors was very low. As for the distribution of oblique case forms, genitive and accusative are also the most frequent in adult speech, and dative is the least frequent. Presumably, children use dative more actively because it is associated with the semantic roles of benefactor/addressee and experiencer that are important in child language.
Babyonyshev’s study relies on the generative framework. For her, the crucial distinction was between structural, inherent, and lexical cases, i.e. the case assigned to an element in a particular syntactic position (e.g. accusative assigned to direct objects of verbs, genitive assigned to complements of nouns), the case assigned to an element bearing a particular semantic role (e.g. dative assigned to benefactors and experiencers), and the case assigned by a particular lexical item (e.g. instrumental assigned to the object of the verb gorditʹsja ‘to be proud of’). However, her predictions were not always borne out. While structural nominative and accusative indeed predated all other case forms, for example, it was impossible to show that structural genitive was easier to acquire or more frequent in child speech than lexical genitive (for an overview, see also Peirce, 2013).
Other studies have addressed different aspects of case-system acquisition by monolingual Russian children (e.g. Cejtlin, 2003, 2009a, 2009b; Ionova, 2007; Popova, 1958/1973; Schütze, 1995; Serebrennikova, 1954; Sizova, 2009; Zakharova, 1958/1973). The study by Janssen (2016) provided the first experimental data on the acquisition of case and gender by monolingual Polish and Russian children and Dutch–Russian and Dutch–Polish bilinguals. 3 In particular, Janssen focused on the phonetic realization of unstressed vowels: in Russian, they are phonetically reduced, while in Polish there is no reduction. All Polish words have penultimate stress, so all inflections are unstressed, but not reduced. In Russian, words have different stress patterns, and stressed and unstressed inflections sound very differently as a result of phonetic reduction. Janssen referred to this to explain her finding that Russian-speaking children acquire cases and genders slower than their Polish peers.
Monolinguals outperformed bilinguals of the same age in all production tasks. Janssen also noted that monolinguals and bilinguals produced different types of case error. Bilinguals often used nominative forms where another case was required; monolinguals did not show this tendency, although they sometimes mixed different oblique cases. Non-experimental studies of children acquiring Russian, both monolingual and bilingual (e.g. Peeters-Podgaevskaja, 2008; Schwartz and Minkov, 2014, as well as the works mentioned above), confirm this generalization. In many unbalanced bilinguals and adult heritage speakers of Russian, the case system is very poor, and the overuse of nominative forms is even more pronounced than in balanced bilinguals. English–Russian heritage speakers studied by Polinsky (2007a, 2007b) actively used only nominative and accusative; genitive was retained in some contexts, while all other cases were used only sporadically.
3 L2 acquisition of cases in Russian and other morphologically rich languages
Now let us turn to case acquisition in Russian as an L2. The only study addressing the acquisition of the whole Russian case system is by Rubinstein (1995a, 1995b). Rubinstein recruited 136 speakers of American English who were learning Russian at the Defense Language Institute in Monterey, California. Seventy-three students were interviewed in the middle of an intensive course (weeks 21–22), while the other 63 students were interviewed at the end of this course (weeks 41–44). Oral interviews consisted of 50 questions prompting participants to use particular oblique cases (10 questions per case).
Rubinstein (1995a) analysed which oblique cases were produced correctly more often, comparing the number of contexts in which a certain case was required to the number of correct forms. Obviously, these contexts were not limited to answers to the 10 questions per case mentioned above, for example, accusative is required in any sentence with a direct object. The questions were designed only to make sure that participants attempted to produce every oblique case. On average, 60% and 74% oblique case forms were produced correctly in the two groups. Participants in both groups made the fewest errors with accusative and locative, then came genitive and instrumental, and dative was the most difficult. This order does not coincide with the order of acquisition by Russian children and differs from the order of presentation in class. Rubinstein argued that his findings could be explained only by a combination of factors: morphological and semantic complexity of different cases, their relative frequency, and the order of presentation in class.
Rubinstein (1995b) did not analyse correct case forms and paid closer attention to errors. In total, 2,570 and 1,705 errors were recorded in the mid-course and end-course groups, respectively. Rubinstein divided them into three categories:
Primary Form, or PF: ‘replacement of the correct form by the primary form (i.e. nominative)’: 1,684 (66%) and 1,069 (63%) errors in the two groups;
Outside-the-Case, or OC: ‘misused inflection between different cases’ (when different oblique cases were mixed): 678 (26%) and 488 (29%) errors;
Within-the-Case, or WC: ‘misused inflection within one case’ (when the student determined the case correctly, but used a wrong declension): 208 (8%) and 148 (9%) errors.
As we showed in the previous section, the same types of error can be found in L1 acquisition, but their distribution is quite different. Using endings from a wrong declension (WC) is the most frequent error, while using nominative where an oblique case is required (PF) is extremely rare (in some datasets, this was not observed at all, while in the others, singular errors were recorded at early acquisition stages). Rubinstein also noted some characteristic case-mixing patterns (OC). For example, in Russian the locative is used to denote place and the accusative to denote direction, and students often confused them (mostly substituting locative for accusative). These errors are also mentioned in other studies on learning Russian as an L2, but not as an L1. In general, Rubinstein concluded that over time, the quantity of case errors changed, but their quality (the distribution of error types) did not.
Other studies dedicated to the L2 acquisition of Russian cases were experimental and looked at comprehension (Gor et al., 2017, 2019; Kempe and MacWhinney, 1998, 1999). They did not compare different oblique cases and analysed such pairs as nominative vs. accusative, nominative vs. genitive, etc. Kempe and MacWhinney (1998, 1999) compared English speakers learning Russian and German: how successfully they processed nominative and accusative forms depending on the word order in the sentence, the animacy of the nouns, etc. Gor et al.’s (2017, 2019) studies mainly focused on morphology, finding out whether L2 learners decompose noun forms into an inflection and a stem during processing, how form frequency affects processing, etc.
If we turn to L2 research on other languages with rich case systems, the existing literature is also rather limited. Several studies looked at Finnish, which has 15 cases (Bellucci and Dal Pozzo, 2016; Kaivapalu and Martin, 2007; Vainio et al., 2016), and at Korean, which has eight cases (Brown and Iwasaki, 2013). They mainly focused on the influence of L1: whether L1 marks cases at all and whether the case system of L1 is similar to that of L2. The results are contradictory: having a (similar) case system in L1 was helpful for the learners in some tasks, but not in others. Comparing different oblique cases was not the purpose of these studies. In other words, there are very few studies on L2 acquisition of the case system as a whole, either in Russian or cross-linguistically. The present study aims to fill this gap, identifying the role of different factors in this complex process.
4 The present study
Like Rubinstein (1995a, 1995b), we aimed to study the L2 acquisition of the Russian case system, but our study differs from his in two important aspects. First, while Rubinstein assessed students’ performance in the middle of and after a very intensive course, our participants went to regular long-term language courses and had six different proficiency levels, which gave us more opportunities to observe the gradual acquisition of the case system and to find not only quantitative, but also qualitative changes.
Second, Rubinstein elicited oral answers to questions, while our participants wrote short texts describing a comic strip. The former method has the obvious advantage of immediacy and allows analysing such features as stress, although Rubinstein did not do so in his papers. The latter method allows analysis of orthography and also gives participants the opportunity to plan more carefully. In particular, they can avoid forms and constructions they are not sure about. We believe that avoidance is a crucial strategy in case of uncertainty, and tracing which cases the participants attempt to produce or try to avoid at a certain level is an important source of information about their internal grammars, on a par with case errors they make. 4
In our study, we relied on these two types of data, while Rubinstein analysed only error rates: which cases his participants attempted to produce was predetermined by the questions he asked. Comparing different oblique cases, Rubinstein used the term ‘accuracy order’ instead of ‘acquisition order’, assuming that his participants most probably would never acquire the Russian case system fully, and that some cases would always remain more difficult than others, triggering more errors. We cannot adopt his term because, for us, error rates are not the only relevant measure; therefore, we use the term ‘acquisition order’ with certain reservations, as well as ‘acquisition trajectories’. This way we can discuss what happens at different acquisition stages.
Analysing our findings, we consider a variety of factors, including the order of presentation in class, the morphological and semantic complexity of different cases (number of functions, variability of inflections), and the range of contexts in which they are required. The last correlates with the relative frequency of different cases. Comparing the resulting picture with L1 studies lets us draw more general conclusions about the different roles of these factors in the L1 and L2 acquisition.
Finally, let us note one more difference between Rubinstein’s and our studies. The questions used by Rubinstein often gave participants a hint of which case they should choose: for example, the instrumental case of the wh-word in (2) (Rubinstein, 1995a: 31, example 6) indicates that the instrumental should also be used in the answer. Of course, a certain proficiency level is necessary to decipher these hints and to put the noun in the answer in the right case, so the task is far from trivial for an L2 learner. Still, our participants had no such hints guiding their choices.
(2) S kem vy xotite tuda poexatʹ? with whomINS you want there to go ‘With whom would you like to go there?’
II Method
1 Participants
The participants in our study were 192 adult learners of Russian as an L2 (109 females, age range 18–65). All studied at the Official Language School Barcelona Drassanes (EOIBD). The testing sessions were organized at the end of the 2013 and 2014 academic years (Denissenko Denissenko, 2016). All students were Catalan–Spanish bilinguals (there were very few speakers of other languages at the EOIBD, so we could not licitly compare several L1s). 5 All participants provided informed consent.
At the EOIBD, a student is placed in a proficiency level group based on the results of an exam at the end of every academic year (it should therefore take one year to pass each level). The levels correspond to the Common European Framework of Reference for Languages (CEFRL). Proficiency levels of our participants ranged from A1 (beginners) to C1 (advanced), with six levels in total (A1, A2, B1, B1+, B2, C1). We recruited 32 participants at each level. Table 4 provides more detailed information about the six participant groups, showing that they are relatively homogeneous in age and gender distribution. The curriculum at the EOIBD is not specifically aimed at university-age students, and the majority of our participants were working professionals, with university students being the second largest group.
Age and gender in the six L2 participant groups.
The students had 150 hours of formal instruction a year, two classes a week. When starting to learn cases, they used the coursebook (Gorbatkina et al., 2003, 2004). This book introduces all cases at the A1–A2 levels, although not in all their functions, which are learned gradually till the B2 level. Cases are presented in the following order: nominative > locative > accusative > genitive > instrumental > dative.
We also recruited a control group of 32 native Russian speakers (18 females, age range 19–65 years, mean age 28). All of them were monolingual and lived in Saint Petersburg, Russia. They were university students and working professionals.
2 Procedure
We aimed to elicit from our participants short written texts that could be compared easily. To do so, we gave them a comic strip from the textbook by Krylova and Sokolova (1988), shown in Figure 1, and the following instructions in Russian, Catalan, and Spanish (we translate them here into English): Task 1: Write a story using the pictures given. It is important to write a coherent text describing the situations in each picture. Divide your text into paragraphs. You have 30 minutes. Task 2: Come up with the end of the story you wrote for Task 1. Write a paragraph with at least five sentences. You have 10 minutes.
The comic strip used to elicit texts.
For native speakers, the instructions were modified. Instead of giving them 30+10 minutes, we asked them to write about 200 words for the first task and about 7 sentences for the second. This way, we obtained texts of roughly the same length as in the C1 group of L2 participants.
3 Data analysis
We made the following preliminary analysis of the data. The 192 texts we elicited were converted into an electronic database that contains 26,914 words. The database does not include any personal information about the participants, only the numeric IDs assigned to them, their age, gender, and proficiency level.
For every noun included in the database, we assessed whether it was in the correct case form or not, including both the choice of the right case and the right inflection depending on the inflectional class. 6 As a result, we had 5,980 correct case forms and 822 errors for further analysis. For every correct form, we noted its case. We then distributed the errors into three groups, analogous to those reported by Rubinstein (1995b).
In the first group, termed Primary Form or PF errors, nominative forms used instead of oblique case forms were included, as in (3). All examples here and below are taken from our database, and the proficiency level of the person who produced them is indicated in parentheses. The second group, termed Outside-the-Case or OC errors, contains incorrectly used oblique case forms, as in (4). In the third group, termed Within-the-Case or WC errors, we included examples in which learners selected an oblique case correctly, but chose an ending from a wrong inflectional class or subclass, as in (5). For example, in (5) the ending for palatalized stem-final consonants was selected instead of the ending for non-palatalized ones. For PF and OC errors, we marked both the case that the learners used and the case that was required. For WC errors, we noted the case for which a wrong inflection was chosen.
(3) PF: primary form (A2 level) *On el ryb he ate fishNOM.SG with tomatoINS.SG ‘He ate fish with tomato.’ (4) OC: wrong oblique case form (B1+ level) *Druzʹja vyxodili na ulitc friendsNOM.PL went out to streetLOC.SG ‘The friends were going out into the street.’ (5) WC: wrong inflectional class or subclass (A2 level) *Oni čitajut žurnal Correct form: žurnal they read magazinesACC.PL and booksACC.PL ‘They are reading magazines and books.’
There were no errors in inflections that could be classified as purely orthographic. Other errors (misspelled noun stems, inappropriate word selection, etc.) were not considered in the present study. In general, the choice of the comic strip to elicit texts was fortunate, because in most cases, we could easily determine what the students wanted to convey, even when they made multiple lexical and grammatical errors at the beginner level. However, several examples were difficult to analyse. For example, in (6) it is not clear what the form škole ‘schoolDAT/LOC.SG’ means. Maybe the student wanted to write ‘in school’, but forgot the preposition; however, even in this case the meaning of the sentence cannot be related to what we see in Figure 1 in an obvious way. We did not want to make any unfounded guesses and excluded such examples from further analysis. Only 10 examples were excluded.
(6) *Ego otec sprava čitaet škole kartu. (A1 level) his fatherNOM.SG on the right reads schoolDAT/LOC.SG mapACC.SG
In the control group of native speakers, we had 1,472 noun forms with no case errors.
III Results
1 Overall picture
First of all, we counted the number of errors and correct case forms at every proficiency level. Table 5 presents the results for all cases and for oblique cases only (excluding nominative).
Correct noun forms and errors at every proficiency level.
Most L2 studies exclusively focus on errors, such as that of Rubinstein (1995b), who followed the principles of the influential Error Analysis framework (Corder, 1981). Table 5 shows that this can be misleading. The number of errors increases from the A1 level to the A2 and then to the B1; it may look like the students made no progress. However, the number of correct forms grows as well, as participants’ texts become longer. If the percentage of errors from all used noun forms is considered, both for all cases and for oblique cases only, we can see that it steadily decreases from level to level.
If all noun forms are considered, the share of errors is already relatively low at the A1 level (15%) and goes down slowly to 8% at the C1 level. But this is primarily due to the contexts in which nominative is required; if they are excluded, the error rate becomes much higher and declines more sharply, from 36% at the A1 level to 12% at the C1 level. Using binary logistic regression, we estimated that this decrease is statistically significant in both cases (β = −0.15, SE = 0.02, z = 47.79, p < .01 for all noun forms; β = −0.25, SE = 0.03, z = 102.01, p < .01 for oblique case forms). 7
In the following sections, we will first look at the different error types introduced above – PF, OC, and WC – and will compare our results to Rubinstein’s (1995b), where he observed that the distribution of these types did not change in his mid-course and end-course participant groups. Second, we will analyse the acquisition trajectories of different cases looking at correct case forms, then at errors, and finally synthesizing the results. Our conclusions will be compared to Rubinstein’s (1995a).
Analysing the overall picture, we did not find any nouns with non-existent endings: participants always substituted for the required inflection some other inflection that is used in another case or declension in Russian. The absence of purely orthographic errors (e.g. confusing inflections that become phonologically undistinguishable due to unstressed vowel reduction) is also notable. Vowel reduction was shown to cause various problems in L1 acquisition, both in oral production and later in writing. The fact that no similar errors were found in our database stresses the difference between L1 and L2 acquisition: our participants acquired cases relying primarily on written examples.
2 Different types of error
Table 6 presents three types of error that we identified, following Rubinstein (1995b): overusing the nominative, or the primary form (PF); using a wrong oblique case (OC); and using endings from a wrong inflectional (sub)class (WC). Whether we consider all noun forms or oblique case forms only, the share of PF errors – those that are most characteristic for L2 learners, but are very infrequent in L1 acquisition – gradually decreases (β = −0.42, SE = 0.04, z = 125.12, p < .01; β = −0.54, SE = 0.04, z = 176.74, p < .01). The number and proportion of OC errors increases from the A1 level to A2 and B1 and then gradually decreases. As we will show below, A2 and B1 are the levels at which our participants started actively using all oblique cases. The share of WC errors is relatively small at all levels, especially for A1 students, who mostly limit themselves to nominative and locative singular forms (the absolute majority of Russian nouns have the same ending -e in locative singular, which precludes WC errors). In L1 acquisition, WC is the most frequent type of case errors, and children also make some OC errors at early stages.
Case errors of different types (percentages from all case forms and from all oblique case forms used at a given level).
Notes. PF = primary form. OC = Outside-the-Case errors (wrong oblique case). WC = Within-the-Case errors (wrong inflectional (sub)class).
Rubinstein (1995b) analysed only oblique case forms and calculated the percentages of PF, OC, and WC errors not from the total number of forms, as we did in Table 6, but from the total number of errors. We provide analogous counts in Table 7. 8 Rubinstein (1995b) observed that the distribution of errors was virtually the same in the mid-course and end-course groups: 63%–66% of PF errors, 26%–29% of OC errors, and 8%–9% of WC errors. He concluded: “Classroom learners move toward native speakers’ competence in the domain of noun cases by decreasing the quantity of their errors rather than by changing their quality” (Rubinstein, 1995b: 426). Our results in Table 7 are very different. The share of PF errors decreases dramatically from the A1 to the C1 level: from 82% to 14% (β = −0.42, SE = 0.04, z = 126.01, p < .01). At the same time, the share of WC errors increases from 7% to 37% (β = 0.15, SE = 0.05, z = 8.23, p < .01). As Table 6 shows, this happens not because their own incidence grows, but because students make fewer other errors. The share of OC errors peaks at the B1 level. Thus, unlike Rubinstein, we found not only quantitative, but also very noticeable qualitative changes.
Oblique case errors of different types (percentages from all oblique case errors at a given level).
Notes. PF = primary form. OC = Outside-the-Case errors (wrong oblique case). WC = Within-the-Case errors (wrong inflectional (sub)class).
3 Correct case forms
To investigate the acquisition trajectories of different cases, we started by analysing correct noun forms at every proficiency level. Their distribution is presented in Table 8. The last row contains the results of the native speakers.
The distribution of correct case forms (percentages from all correct noun forms used at a given level).
Notes. NS = native speakers.
Let us briefly comment on the native speakers’ results first. The observed distribution of noun forms is similar to those found in the corpus-based studies discussed in the introduction. The only difference is that genitive was more frequent in these studies, outnumbering accusative and sometimes also nominative (see Kopotev, 2008). This can be explained by the fact that the share of genitive forms depends on the syntactic complexity of the text (e.g. genitive is used in complex noun phrases, with the majority of prepositions, especially with morphologically complex ones), and describing a comic strip does not prompt the use of particularly complex syntactic constructions.
As for L2 learners, at the A1 level they produced many correct nominative forms, as well as some locative and a smaller number of accusative forms, while other case forms were virtually absent. At the A2 level, the share of accusative forms increases significantly (β = −0.81, SE = 0.15, z = 29.55, p < .01), and so does the share of genitive and instrumental forms (β = −1.16, SE = 0.30, z = 14.98, p < .01; β = −1.84, SE = 0.34, z = 28.55, p < .01). 9 The share of dative forms is still very small, but it grows significantly at the B1 level (β = −1.77, SE = 0.38, z = 21.28, p < .01). After that, there are no significant differences between levels for any case. Only the share of genitive forms keeps growing slowly, resulting in a significant difference between the B1 vs. the C1 level (β = −0.67, SE = 0.15, z = 21.12, p < .01).
The resulting sequence of cases (nominative > locative > accusative > genitive and instrumental > dative) is similar to the findings reported by Rubinstein (1995a), but we will postpone the discussion of this to the following sections, where case errors are analysed. Now let us only note that, unlike Rubinstein, we see a clear difference between accusative and locative, and it cannot be simply reduced to the order of acquisition. Locative is acquired earlier, but is outnumbered by accusative forms at subsequent levels.
The overall distribution of cases in the texts changes as a result of case-system development. The share of nominative forms gradually decreases from 70% to 42% at the advanced level. The share of locative forms decreases from 18% to 11%–12%. As Table 8 shows, the absolute number of locative forms hardly changes; the decrease is primarily due to the fact that students used more forms of other cases, not fewer locative forms. The proportion of other cases slowly increases: genitive from 2% to 14%, dative from less than 1% to 4%, accusative from 9% to 20%, instrumental from 1% to 8%. As a result, at the advanced level, the distribution is similar to that in native speakers’ texts, although the share of nominative is still considerably larger (42% vs. 35% for native speakers).
4 Errors with different cases
Table 9 and Figure 2 present the numbers and percentages of errors in the contexts in which a given case is required. As with the total number of errors, if we look at raw numbers, it may seem that the learners’ knowledge of certain cases degrades in the course of their studies. For example, the number of errors in the contexts where dative is required increases from level A1 to A2 and then to B1; the number of errors in the contexts where genitive is required increases from level A2 to B1 and then to B1+. But, since the number of correct forms grows as well, the share of errors decreases steadily for all oblique cases, with one notable exception, which we will discuss below: the percentage of errors in the contexts where locative is required increases from level A1 (8%) to level A2 (19%).
The distribution of errors (percentages from all contexts in which a given case was required). 10
The distribution of errors.
The overall decrease in error rates is significant for genitive (β = −0.41, SE = 0.07, z = 38.17, p < .01), dative (β = −0.29, SE = 0.09, z = 9.55, p < .01) accusative (β = −0.39, SE = 0.05, z = 76.17, p < .01), and instrumental (β = −0.35, SE = 0.06, z = 31.34, p < .01). The results for locative are significant only if the A1 level is excluded (β = −0.22, SE = 0.08, z = 7.52, p < .01). The error rate in the contexts in which nominative is required was 1%–3% at all levels. In the next section, we will compare error rates for different oblique cases to draw conclusions about the order of acquisition.
Now, let us look at locative. It is the first oblique case to be learned, but the detailed analysis of our data demonstrates that initially, it is acquired only in one of its functions: with two prepositions indicating location, v ‘in’ and na ‘on’. At the A1 level, participants do not even try to use locative with other prepositions and in other contexts (to indicate the object of speech and thoughts, time or means of transport) and rarely use any other prepositions requiring different cases. All 146 correct locative forms are used with v ‘in’ and na ‘on’, and the few errors in the contexts where locative was required (11 out of 13) are also made with these prepositions: nominative was used instead of locative.
Starting from level A2, other errors appear. Learners start using more prepositions that require accusative, instrumental, and genitive; sometimes they use these cases erroneously instead of locative (11 out of 30 errors in the contexts where locative was required). Additionally, the preposition o ‘about’, which requires locative, is particularly difficult for the students (maybe because it does not have a locative meaning): 12 errors involved using nominative with this preposition. There were only two cases out of 30 in which nominative was used instead of locative with v ‘in’ or na ‘on’. Thus, even this case, when examined in more detail, illustrates the progress of case system acquisition.
IV Discussion
In this section, we will first trace the acquisition trajectories of different cases, summarizing the results and comparing them to Rubinstein’s (1995a). Second, we will look at the results from a different perspective, discussing the maturational changes at different proficiency levels. Finally, we will turn to factors that may play a role in case-system acquisition.
Let us start by comparing different cases. Rubinstein registered the following error rates in his mid-course and end-course groups, respectively: 33% and 21% for accusative, 31% and 23% for locative, 41% and 29% for genitive, 45% and 29% for instrumental, and 56% and 45% for dative. Based on the statistical analysis, he established the following accuracy order for oblique cases: first accusative and locative, then genitive and instrumental, and finally dative. As we mentioned in the introduction, Rubinstein used this term to reflect his conclusion that some cases would always remain more difficult for L2 learners, triggering more errors than others. Thus, as in the distribution of different error types, he observed quantitative changes (decreasing error rates for every oblique case), but no qualitative changes here.
Our results do not contradict the order established by Rubinstein, but allow for a much more fine-grained picture, including qualitative changes. According to our data, locative is acquired earlier than accusative, although initially only in its most important function: with prepositions indicating location. Accuracy rates for locative and accusative differ significantly at the A1 level (β = −2.31, SE = 0.33, z = 47.69, p < .01) and the B1 level (β = −2.31, SE = 0.33, z = 47.69, p < .01), but not at the A2 level, when students start acquiring other functions of locative, as well as various prepositions requiring different cases. At the same time, since accusative is really essential as the direct object case, it already outnumbers locative at the A2 level (163 vs. 132 correct forms, 215 vs. 162 contexts where these cases are required).
If we look at the proportions of errors, genitive is similar to accusative, except for the B1+ level, where they differ significantly (β = −0.64, SE = 0.28, z = 5.16, p = .02). At this level, the error rate for accusative falls by 10%, while the error rate for genitive remains the same, to fall at the subsequent levels. However, the numbers of correct accusative and genitive forms and the numbers of contexts where these cases are required reveal very different acquisition trajectories. Students use many accusative forms already at the A1 level; their share increases at the A2 level and then remains relatively constant. Conversely, the share of genitive forms gradually grows from the A1 level to the C1 level.
The error rates for instrumental are always higher than for genitive, although this difference reaches statistical significance only at the C1 level (β = −0.97, SE = 0.38, z = 6.67, p = .01). 11 The difference in error rates across all levels is also significant (β = −0.46, SE = 0.14, z = 11.53, p < .01). As for correct forms, initially their numbers are similar, but the share of instrumental forms plateaus at 7%–8% at the A2 level, while the share of genitive forms keeps growing. Thus, we can conclude that genitive is acquired earlier than instrumental, but its acquisition trajectory is more complex as a whole, presumably because it has the largest number of functions that are acquired gradually. 12
Dative consistently has the highest error rates and the lowest number of correct forms of all the cases. If we compare dative to instrumental, which has the second highest error rates, the difference is statistically significant across all levels (β = −0.34, SE = 0.16, z = 4.73, p = .03). 13 The problems L2 learners experience with dative are in sharp contrast with L1 acquisition: as we showed in the introduction, in children’s speech, dative forms are more frequent than locative and instrumental.
Notably, our data do not support Rubinstein’s (1995a) claim that the accuracy order remains constant in the course of acquisition. At the A1 level, the error rates for locative are much lower than for accusative and especially for genitive. At the C1 level, their error rates converge (the error rate for genitive is even slightly lower than the error rates for locative and accusative). Starting from the work of Ellis (1994), the notions of order and sequence have been central for L2 acquisition studies. In particular, Ellis demonstrated that grammatical categories and inflectional morphemes in L2 English are acquired in a particular order, independently from the learner’s L1. Our data show that the initial order of acquisition is only part of the picture, which does not fully predetermine the dynamics of the system at later acquisition stages, contrary to Rubinstein’s conclusions.
To summarize our findings from a different perspective, let us sketch what happens at different levels. At the A1 level, we recorded 575 correct nominative and 146 correct locative forms. The error rates for these cases are very low. Students try to avoid all other cases. To do so, they limit themselves to very simple sentences, like those in (7a–b).
(7) a. Studenty v biblioteke. (A1 level) studentsNOM.PL in libraryLOC.SG ‘The students are in the library.’
14
b. Ivan i Anna čitajut. (A1 level) IvanNOM.SG and AnnaNOM.SG read ‘Ivan and Anna are reading.’
Direct objects are difficult to avoid, so there are also 76 correct accusative forms and 68 errors in the contexts where accusative is required. The contexts in which genitive, instrumental, and dative are required are limited to 34, 40, and 9, respectively. The error rates are very high: 56%, 75%, and 89%.
At the A2 level, the contexts in which genitive and instrumental are required grow dramatically (to 67 and 90, respectively), while the error rates fall sharply (to 27% and 30%). The number of contexts requiring accusative also grows (to 215), and the error rate decreases. The same happens to dative at the B1 level: the number of contexts grows from 21 to 77, and the error rate falls from 62% to 34%. Thus, when learners feel more confident, they start using these cases more actively. The share of accusative, instrumental, dative, and locative forms remains constant at the subsequent levels, while the share of genitive forms gradually increases.
Now let us turn to factors that may play a role in case-system development. The acquisition order we observed coincides with the order of presentation in the coursebook used by our participants: locative > accusative > genitive > instrumental > dative. However, reducing everything to this factor would be overly simplistic. First, dative was the most difficult also for the participants of Rubinstein’s (1995a) study, although in their coursebook it was introduced before instrumental. Second, all cases are introduced at the A1–A2 levels, while the differences between them can still be observed much later in the course of acquisition, which can hardly be explained exclusively by the initial order of presentation. Third, and most importantly, the order of presentation in different coursebooks is not random, but is based on teaching experience: which cases are more necessary, and which are more difficult? We hypothesize that it is the interplay between these two groups of factors – for the sake of brevity, we will refer to them as ‘necessity’ and ‘complexity’ – that defines the order of presentation, the order of acquisition, and subsequent acquisition trajectories.
Necessity depends on two things: the range of contexts in which a given case is required, and how essential these contexts are for successful language use. To illustrate the difference between the two, let us look at genitive and accusative. Genitive is required in the widest range of contexts; it is the most frequent oblique case according to different corpus counts, as we showed in the introduction. Accusative covers a smaller variety of contexts, but one of its functions, encoding direct objects, is absolutely indispensable. This is why in the present study, accusative is the most frequent oblique case starting from the A2 level, while genitive lags behind at the early stages, but eventually catches up at the C1 level, becoming the second most frequent oblique case.
Complexity includes morphological complexity (variability of inflections and syncretism) and semantic complexity (the number of case functions). Thus, there is a correlation between these factors: the more functions a case has, the more often it is required.
In terms of necessity, accusative and genitive should come first. Accusative serves the vital function of direct object, which is why it is the first oblique case to appear in L1 acquisition. Genitive covers the widest array of functions. Indeed, accusative is the first oblique case in L1 acquisition, and genitive comes second.
However, at the early stages, L2 acquisition of accusative is impeded by its morphological complexity. Depending on the inflectional class and the animacy of the noun, the ending of the accusative may coincide with the nominative or the genitive, or be unique (see Tables 1 and 2). Later, learners experience problems mastering different functions of accusative. In particular, they often confuse accusative used to denote direction, with locative denoting place. Genitive is semantically complex due to its high number of functions and is especially morphologically complex in plural, with the largest inflection variability in the whole Russian case system. This is why in L2 acquisition, accusative and genitive are preceded by locative in its most basic function. Locative has the lowest morphological complexity: the majority of Russian nouns have the inflection -e in locative singular.
These three cases are followed by instrumental and then dative. Instrumental and dative are the least frequent according to different corpus studies, although locative is not very different from them (see Table 3). Dative serves the functions of benefactor/addressee and experiencer (especially with impersonal predicates). These functions are crucial in child language, so dative appears early during L1 acquisition, but are apparently are not so important for adult native speakers and L2 learners. Other functions of dative seem to be particularly difficult for L2 learners, because they provoke confusion. For example, dative is used with the prepositions po ‘along, according to’ and k ‘to’. The former has a plethora of meanings. The latter is used primarily to indicate direction, but accusative is more salient in this function, and locative is more salient as a case with a locative meaning in general. Morphological syncretism complicates the situation even further: the dative singular inflection coincides with the locative singular one for some nouns and with the accusative singular inflection from a different declension for the others (see Tables 1 and 2).
Instrumental, on the contrary, has very distinctive endings (the only two-letter endings in singular paradigms). The array of functions served by instrumental is diverse and complicated. However, our participants mostly used it with the preposition s ‘with’, which has a simple meaning. As a result, instrumental was acquired faster than dative.
Summarizing the results, let us compare L2 and L1 case-system acquisition. First, we have already noted that the type of case errors that is the most characteristic for L1 (WC errors) is the least characteristic for L2. The opposite is true for PF errors. However, contrary to Rubinstein’s (1995b) conclusions, the distribution of error types changes as the L2 system matures, becoming more similar to L1.
Second, we can conclude that both necessity and complexity play a major role in L2 acquisition. In L1 acquisition, the balance is shifted in favor of the former: accusative, genitive, and dative are more frequent than locative and instrumental at the initial stages. The crucial role of morphological complexity provides evidence in favor of the theories of second language acquisition (SLA) arguing that morphology is the main source of difficulties for L2 learners, and thus the major source of differences between the L1 and L2 (e.g. Slabakova, 2009, 2014). In the L2 processing literature, there are also a number of studies regarding inflectional morphology as the weak link (e.g. McCarthy, 2008; Prévost and White, 2000).
Analysing the errors, we also found that very few OC errors are completely random. One oblique case is used instead of another when they have similar functions or based on analogy with a synonymous verb. This shows that despite various difficulties, cases are acquired as a system, and that many generalizations are present in the mental grammar of L2 learners.
V Conclusions
Using texts elicited from students with different proficiency levels, we studied nominal cases in Russian as an L2. The maturation of the system is evident both in the number of correct forms participants produced at different levels and in the error rates. Cases were acquired in the following order: nominative, locative (although initially it is actively used only in one of its functions: denoting place), accusative, genitive, instrumental, and dative. The acquisition trajectory of genitive is the most complex since it has the highest number of functions: the simplest functions are mastered early, but more functions appear at later stages, and the number and share of genitive forms constantly grows until the most advanced levels.
We have argued that case-acquisition trajectories are determined by two groups of factors: how essential a given case is for successful language use (this depends on its functions and is reflected in its relative frequency) and how complex it is, both semantically and morphologically. In L1 acquisition, the role of complexity appears much less significant. As a result, the order of acquisition is different.
Further research is necessary to test and elaborate our hypothesis. First, different factors may be teased apart in analysing case for pronouns and adjectives, as well as case for nouns depending on their number, gender, animacy, and inflectional class. Case functions would remain the same, while morphological characteristics (syncretism and variability of inflections) would vary. Second, it would be useful to study different case functions in more detail.
It should also be noted that we registered very few completely random case errors and no noun forms with non-existent inflections. This shows that cases are acquired as a system from the very start. Furthermore, we found no inflections with purely orthographic errors. Such errors triggered by phonetic reduction are very common for native speakers, especially during their preschool and elementary school years. The absence of such errors in our data can be explained by the fact that our participants acquired paradigms relying primarily on written examples.
The only previous study addressing L2 acquisition of the Russian case system as a whole was conducted by Rubinstein (1995a, 1995b). He relied on oral interviews with students in the middle and at the end of an intensive language course and established the following order of cases: accusative and locative, then genitive and instrumental, and finally dative. Our study gave a more detailed picture, including some characteristics of the acquisition trajectories of different cases. This has not been previously done either for Russian or for other languages with rich case systems.
The majority of L2 studies focus on errors, while our study demonstrates how crucial it is to take correct forms into account. In particular, we found that if a certain error type is rare, it does not mean that the relevant case has been successfully acquired. At early stages, this means that learners do not attempt to use verbs and prepositions requiring this case. When they start trying more actively, the raw number of errors may initially increase together with the number of successful attempts. Therefore, only the distribution of correct and incorrect forms taken together can be meaningfully interpreted.
In the course of acquisition, the distribution of correct case forms changes, gradually approaching that observed in native speakers. The distribution of errors also undergoes important changes. The share of errors that are the least characteristic for L1 acquisition (using nominative forms instead of oblique case forms) declines dramatically, while the share of errors that are the most characteristic for L1 acquisition (using an inflection from a wrong inflectional [sub]class) increases. This is another major difference between Rubinstein’s (1995a, 1995b) and our studies: he found only quantitative changes (the incidence of different errors decreased, but their distribution remained the same), while we also observed qualitative changes. Nevertheless, our most advanced participants are still far from native speakers, and they make many errors, because case is one of the most complex morphological categories that is very difficult for L2 learners.
Footnotes
Acknowledgements
We are especially grateful to Nina Avrova Raaben, Olga Leontieva, and Oksana Norko, the teaching staff of the Russian Department of the Official Language School Barcelona Drassanes (EOIBD), Spain, for their help with data collection.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Natalia Slioussar’s work on this project was carried out in the framework of the Basic Research Program at the National Research University Higher School of Economics, Russia.
