1. Introduction

A major question in the field of second language acquisition is to what extent interlanguage variability stems from, on the one hand, first-language (L1) influence and, on the other hand, the nature of the second language (L2) itself. While variability in Bates and MacWhinney’s (1981) Competition Model is the result of competition between L1 and L2 structures, Ellis (2002) emphasises the role of input frequency; he describes L2 acquisition as “the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within them” (p. 143, emphasis added). When it comes to L2 acquisition of articles, patterns of omission have often been accounted for in terms of L1 influence but rarely in terms of input frequency (Ogawa, 2015). This is probably because “articles are the most frequent forms that are available to learners in input” (Young, 1996, p. 136). Nevertheless, Ogawa’s corpus study on count and mass nouns revealed that typical patterns of article omission in learners of L2 English reflect to some extent patterns in the input. It is therefore worthwhile to consider the role of probabilistic regularities in the target language itself when seeking to explain patterns of article omission in L2 learners.

The two factors – L1 influence and input frequency – may interact, and disentangling them can be complicated. For example, L2 learners of English whose L1 does not have articles are more likely to omit an article in an adjectivally modified noun phrase (NP) than in a non-modified one (e.g., Pongpairoj, 2007; Snape, 2006; Trenkic, 2000, 2007). This could be for two reasons. One reason is that adjectivally modified NPs are more complex than non-modified ones and therefore require more attentional resources. Because these resources, which are limited, are also needed to inhibit selection of L1 structures, the article-less L1 structure has a greater chance to be selected when the NP includes an adjective than when it does not. The second reason is that adjectivally modified NPs are less frequent in input than non-modified ones, which results in stronger contingency between articles and nouns than between articles and adjectives. These two possibilities are not mutually exclusive (Austin et. al., 2015; Trenkic, 2009).

While the majority of data that lies at heart of different explanations comes from research on L2 English, a language with a relatively simple article system, the present study investigates variability in L2 Swedish, a language that encodes definiteness using a more complex NP structure. Specifically, the study focuses on Russian-speaking learners’ production of Swedish NPs. Russian does not have articles; indefiniteness and definiteness can be expressed indirectly by quantifiers and determiners, which are normally placed at the left edge of the NP, but in their absence, NPs are bare (Sussex & Cubberly, 2006). By contrast, Swedish obligatorily expresses definiteness using both free-standing morphemes at the left edge of the NP and suffixes – inflectional morphemes at the right edge of adjectives and nouns (Teleman et al., 1999). Some of these morphemes are relatively frequent in input while others are less frequent. By exploring this acquisitional challenge, the study aims to shed light on some mechanisms underpinning variability in L2 production of functional morphology.

2. Background

3. The current study

The current study focuses on the production of adjectivally modified and non-modified NPs in indefinite and definite contexts – that is, the four structures exemplified in (2) above – by Russian-speaking learners of Swedish. These learners need to learn that Swedish obligatorily expresses indefiniteness using an article at the left edge of the NP (en) and definiteness using a nominal suffix (-en). They further need to acquire that a definite NP that includes an adjective requires not just the definite nominal suffix but also a free-standing definite article at the left edge of the NP (den) and a definite agreement marker suffixed to the adjective (-a). The study explores this acquisitional challenge by posing the following research question: What patterns of omission of functional morphology can be discerned in Russian-speaking learners’ production of Swedish NPs?

Specifically, because previous research suggests that input frequency may affect L2 acquisition of functional morphology, the analysis first compares the suppliance of high- and low-frequency morphemes. Second, because previous research suggests that the presence of inflectional morphology in the L1 may facilitate the acquisition of L2 suffixes, the analysis compares the production of bound morphemes (suffixes) and free-standing ones (articles). Finally, because previous research discusses the possibility that the negative effect of adjectival modification on article suppliance may be due to strong contingency between articles and nouns, the analysis compares the effect of modification on the indefinite left-edge article – en (vit) bil – and the definite nominal suffix – (den vit-a) bil-en.

4. Method

5. Results

Before comparing the proportions of NPs including the four morphemes, it is worthwhile to describe what the NPs looked like. Tables 4 and 5 give the number of adjectivally non-modified and modified NPs broken down by group, context and structural pattern. When it comes to the L1 speakers, they primarily used the four structures in (2), as expected. Sometimes, however, they used den in non-modified NPs; it then functions as a demonstrative pronoun (e.g., den bil-en ‘that car’). Sometimes they also omitted the definite left-edge article in definite modified NPs (e.g., ?vit-a bil-en ‘the white car’), which is not necessarily ungrammatical (see Section 2.2). A bare NP occurred only once in the L1 data. Moreover, the L1 speakers sometimes used indefinite forms in definite contexts and vice versa. When they overused definite forms, they appeared to conceptualise identical blocks as several instances of the same referent. For example, they might say: Pojken är i parken, och pojken är i skolan också ‘The boy is in the park, and the boy is in the school too’. When they overused indefinite forms, this was because they explained how to place the blocks in an almost over-explicit way. For example, they might say: Du ser en vit bil i parken – sätt en svart katt på den ‘You see a white car in the park – put a black cat on it’ instead of Sätt en svart katt på den vita bilen i parken ‘Put a black cat on the white car in the park’. Nevertheless, the data collected from the L1 speakers suggest that the task was overall successful in eliciting the four target structures in expected contexts.

When it comes to the L2 learners, modified NPs appeared to be more difficult than non-modified ones, and definite NPs appeared to be more difficult than indefinite ones. Indefinite, non-modified NPs were accurately produced in 90.4% of obligatory contexts, definite non-modified NPs in 77.4%, indefinite modified NPs in 61.9%, and definite modified NPs in 27.3% of obligatory contexts. The dominant patterns were, except for the target structures in (2), bare NPs (e.g., *vit bil), definite modified NPs with only the definite nominal suffix (e.g., *vit bil-en), and definite modified NPs without the definite left-edge article (e.g., ?vit-a bil-en). The definite adjectival suffix (-a) almost never occurred without the nominal suffix (-en), and the definite left-edge article (den) almost never occurred without the adjectival suffix (and never without the nominal suffix). In other words, the definite left-edge article implies the definite adjectival suffix, which implies the definite nominal suffix. Moreover, indefinite and definite morphemes rarely co-occurred.

The boxplots in Figure 1 display the individual proportions of NPs that included the indefinite article (en), the definite nominal suffix (-en), the definite adjectival suffix (-a) and the definite left-edge article (den) broken down by group and NP category. A comparison between the two groups with respect to these proportions, summarised in Table 6, confirms that, for each of the four morphemes, the learners appeared to be somewhat more prone than the native speakers to omit it in obligatory contexts (although the difference was not significant for the definite left-edge article in non-modified NPs) and somewhat more likely to supply it in inaccurate contexts (although the difference was not significant for the indefinite article in non-modified NPs and for the definite nominal suffix in modified NPs).

Table 7 shows that the L1 speakers did not generally supply high-frequency morphemes more consistently than low-frequency ones or vice versa, although they produced the high-frequency indefinite article (-en) more consistently in obligatory contexts than the low-frequency definite article (den). It will be recalled that omitting den is not necessarily ungrammatical. By contrast, there was a clear effect of frequency in the L2 group: The learners supplied the two high-frequency morphemes (en and -en) more consistently in obligatory contexts than the two low-frequency ones (den and -a).

Table 8 shows that there were no significant differences between the suppliance of bound and free-standing morphemes in obligatory contexts, neither in the L1 group nor in the L2 group. It is important to note that comparisons were carried out only between the two high-frequency morphemes (en and -en) and between the two low-frequency morphemes (den and -a); low- and high-frequency morphemes were not compared with each other here.

Finally, Table 9 shows that there was no significant effect of adjectival modification in the L1 group. By contrast, in the L2 group, there was a negative effect of modification on suppliance in obligatory contexts of the indefinite article (at the left edge of the NP) but no effect at all on suppliance in obligatory contexts of the definite nominal suffix.

On a final note, it should be pointed out that the three comparisons reported – between high- and low-frequency morphemes, between bound and free-standing morphemes, and between non-modified and modified NPs – took into account suppliance in obligatory contexts only. Previous research has shown that L2 learners from article-less L1 backgrounds generally supply and overgeneralise definite articles to a higher extent than indefinite ones (Goad & White, 2004; Huebner, 1985; Snape 2006; Trenkic, 2000, 2002). Such a general preference for definite forms may admittedly be a confounding factor when comparing the indefinite left-edge article (en) to the definite nominal suffix (-en). In the present dataset, however, the indefinite article was not overused more than the definite nominal suffix or vice versa, neither by the L1 speakers (non-modified NPs: Z = 0.00; p = 1.000; modified NPs: Z = 1.63; p = 0.174) nor by the L2 learners (non-modified NPs: Z = 0.67; p = 0.530; modified NPs: Z = 0.54; p = 0.609). Thus, the learners investigated did not appear to prefer definite forms over indefinite ones in general.

6. Discussion

The study explored variability in definiteness marking in Russian-speaking learners of Swedish. Findings include that (i) high-frequency morphemes were supplied in obligatory contexts more consistently than low-frequency ones; (ii) bound morphemes were not supplied in obligatory contexts more consistently than free-standing ones or vice versa; and (iii) suppliance of the free-standing indefinite article (en) was negatively affected by adjectival modification while there was no such effect for the definite nominal suffix (-en). These observations are now discussed in terms of L1–L2 competition and input frequency.

First, regarding high- and low-frequency morphemes, the study corroborates findings from previous research showing that the Swedish definite modified NP, with its three definite morphemes (e.g., den vit-a bil-en ‘the white car’), poses a particular challenge to L2 learners (cf. Axelsson, 1994; Nordanger, 2017; Nyqvist, 2018). It was produced accurately far less often than the three other target structures, each of which requires only one grammatical morpheme (bil-en ‘the car’; en bil ‘a car’; and en vit bil ‘a white car’). In particular, the two morphemes that are required only in definite modified NPs and which are therefore relatively infrequent (den and -a) were supplied less consistently than the two morphemes that are required regardless of modification and that are therefore relatively frequent (en and -en). This finding is in line with what might be expected on a frequency-based account of L2 acquisition (cf. Ellis, 2002). It is also in line with Ogawa’s (2015) suggestion that patterns of article omission to some extent reflect frequency patterns in the input available to the learners.

Second, regarding bound and free-standing morphology, the learners were equally good at supplying the indefinite left-edge article (en) and the definite nominal suffix (-en), both of which are relatively frequent, in obligatory contexts. Further, the learners did not overuse any of these two morphemes more than the other. These results are unexpected for two reasons. First, previous research suggests that L2 learners of Swedish with a highly inflectional L1, like Russian, are more sensitive to bound than to free-standing morphology (cf. Latomaa, 1992; Nordanger, 2017). Second, previous research also reports that L2 learners from article-less L1 backgrounds generally overuse definite articles more than indefinite ones (Goad & White, 2004; Huebner, 1985; Snape 2006; Trenkic, 2000, 2002). Together, these observations arouse suspicion that the oral-production task utilised to elicit NPs somehow favoured the indefinite article. Further, it should be mentioned that the learners tended to supply the definite adjectival suffix (-a) more consistently than the definite left-edge article (den), although the difference was not statistically significant. Nevertheless, since the same tendency was observed in the native speakers too, the pattern may not necessarily be attributed to the distinction between bound and free-standing morphology but could just as well reflect regularities in the target language itself.

Third, regarding the effect of adjectival modification, only the indefinite article (en) was negatively affected by modification; there was no such effect for the definite nominal suffix (-en). This result may be accounted for with a frequency-based theory (Ellis, 2002; cf. Austin et al., 2015; Trenkic, 2009). In this view, the strong contingencies between the noun and the indefinite article and between the noun and its definite suffix result in the emergence of two deeply entrenched constructions: en N and N-en. However, the former construction is inapplicable when the noun is modified by an adjective, as the adjective intervenes between the article and the noun (en + ADJ + N), while the latter can easily be combined with an adjective (ADJ + N-en). Indeed, the learner data seem to support this idea: The definite nominal suffix (-en) frequently occurred without the definite adjectival suffix (-a) and the definite left-edge article (den), while the definite article and the adjectival suffix practically never occurred without the definite nominal suffix. It certainly appears as if the learners can merge an uninflected adjective with the high-frequent N-en construction, resulting in a structure like *vit bil-en, before a complete representation of the Swedish definite NP structure – den vit-a bil-en – can be constructed.

It is unclear what Austin et al.’s (2015) structural-competition model would predict with regard to Russian-speaking learners’ production of Swedish NPs. According to the model, the more complex the NP is, the more likely it is that functional morphology not provided by the L1 is omitted, everything else being equal. However, when comparing the suppliance of the definite nominal suffix (-en) in adjectivally modified and non-modified NPs respectively, everything else is not equal. On the one hand, it could be hypothesised that the effect of adjectival modification would be worse for the definite nominal suffix (-en) than for the indefinite article (en) because there are three grammatical morphemes competing for attentional resources in definite modified NPs (i.e., den, -a and -en). However, this was not what the data looked like. On the other hand, as pointed out by Danijela Trenkic (personal communication, November 24, 2020), it might be that the three definite morphemes do not compete with each other but, quite the reverse, co-activate each other since they encode the same concept. If this is so, when a learner manages to supress the L1 structure and selects the first definiteness marker (den), the likelihood that the second marker (-a) is also selected increases, and when this happens, the likelihood that the third marker (-en) is also selected increases even more. And this was indeed what was observed in the data: The definite nominal suffix was supplied more consistently than the definite adjectival suffix, which was supplied (slightly) more consistently than the definite left-edge article.

The question thus remains to what extent the negative effect of adjectival modification on article suppliance can be explained in terms of, on the one hand, L1–L2 competition and, on the other hand, frequency-biased regularities in the L2 itself. It is therefore interesting to discuss how well the two approaches elaborated on above account for findings from previous research. First, the frequency-based story, according to which the strong association between the article and the noun is the key factor, cannot explain why Austin et al.’s (2015) Thai-speaking learners of English were more prone to drop the plural -s in more complex NPs, as nothing intervenes between the noun and the plural marker. However, as pointed out by the authors, frequency might still be a contributing factor given the fact that definite plural NPs are less frequent than indefinite plural NPs. Second, an advantage of the frequency-based explanation is that it accounts for the negative effect of adjectival modification on article production observed in children acquiring their L1 (Clahsen, 1994; Granfeldt, 2000) – in this learning situation, there is no competing language to inhibit. Finally, an advantage of the L1–L2 competition explanation is that it accounts for the fact that the negative effect of adjectival modification on article production is typically not found in L2 learners whose L1 does have articles (Granfeldt, 2000; Nordanger, 2017; Pongpairoj, 2007).

In this context, it should be pointed out that Axelsson (1994) noticed a negative effect of adjectival modification on the production of indefinite articles in Spanish-speaking learners of Swedish. This is surprising as Spanish actually has articles (at the left edge of the NP). However, Spanish adjectives are by default post-nominal. Thus, Spanish-speaking learners of Swedish may in principle employ an L1 structure for producing an indefinite non-modified NP (ART + N) but not for producing an indefinite modified NP (Swedish: ART + ADJ + N; Spanish: ART + N + ADJ). Together, these observations suggest that frequency-biased regularities in the target language are crucial when it comes to establishing new constructions. For L1 acquisition, all constructions are new ones, while for L2 acquisition, constructions without direct counterparts in any previously learned language are new ones. This may explain why article production is negatively affected by adjectival modification both in children acquiring their L1 and in L2 learners whose L1 does not have articles. At the same time, it may also explain why the production of the indefinite article but not the definite nominal suffix is negatively by adjectival modification in L2 learners of Swedish, both learners whose L1 does not have articles and learners whose L1 is Spanish.

On a final note, a reason why it is complicated to disentangle L1–L2 competition from regularities in the L2 itself is that the key factors – complexity and frequency – are typically negatively correlated: Frequent phenomena tend to be simple and vice versa (e.g., Goldschneider & DeKeyser 2001: 58). Further research therefore needs to design studies that can tap into these two different but related mechanisms separately.

7. Conclusion

By exploring patterns of omission of L2 functional morphology in Russian-speaking learners’ production of Swedish NPs, this study has contributed to the discussion about L1 influence and input-frequency effects in interlanguage variability. Some observations were readily accounted for in terms of input frequency. Simultaneously, the discussion showed that disentangling the two phenomena is complicated. It can be concluded that, while the effect of L1–L2 competition should by no means be neglected, input frequency is a crucial factor in the acquisition of new constructions.