The Norwegian particles jo and nok in second language writing A qualitative study of three learner groups from the ASK-corpus

This paper deals with the topic of lexical modality in Norwegian as a second language. Basing on data obtained from the ASK-corpus – The Norwegian Language Learner Corpus containing second language texts written in a language examination, the authors analyse the use of two modal particles, jo and nok, by three groups of Norwegian as a second language writers: English, Polish and German. The focus of the study is on analysing lexical patterns for co-occurrence of the modal particles. The patterns used by the learners are compared with the ones used by the native speakers of Norwegian and between the learner groups. The discrepancies found in the data are discussed within the broader framework of second language development and second language writing. The findings suggest that the second language writers’ use of modal particles is influenced by several factors, such as general interlanguage tendencies, transfer from the learners’ first languages and the perception of textual norms.


The scope of the study
This paper focuses on Norwegian modal particles, understood as short uninflectable lexical items that do not contribute to the propositional content of an utterance (Foolen 1996) 1 , but connect the current utterance to a pragmatic context (Fischer 2007). Although modal particles, or rather more generally pragmatic markers in Norwegian have received a considerable amount of attention (as evinced for instance by the special issue of Norsk lingvistisk tidsskrift 36/2), their use by second language (henceforth L2) speakers has so far not been studied in detail (see chapter 2 for a review of literature). In this paper, we focus on two particles jo and nok 2 , and their use by L2-writers with three different first language (henceforth L1) backgrounds, compared with native speakers.
We seek to answer the following research questions: What are the characteristics of the use of jo and nok in Norwegian L1 and L2 writing?; How can the differences in the use of jo and nok by three groups of L2 writers, be accounted for?
There is an agreement that L2 speakers generally use modal particles with a lower frequency than L1 speakers (Möllering 2001: 132). It is however unclear why it is so. One of the possible reasons may be the avoidance of those L2 structures that are different from L1 (Schachter 1974: 212). The tendency of one's L1 to influence L2 is known as transfer which we understand as "the influence of a person's knowledge of one language in that person's knowledge or use of another language" (Jarvis -Pavlenko 2008: 1). Odlin (1989: 27) says that this influence results from similarities and differences between the two languages in question, which do not need to be objective. Kellerman (1979) proved that learners make judgements on what is transferable or not, based on their actual knowledge of the L2 and their perceived distance between the L1 and the L2. In language typology, modal particles are seen as a phenomenon typical for the Germanic languages, with the notable exception of English (Harbert 2007: 32-35). Polish, on the other hand, belongs to the Slavonic language family, which in the studies of pragmatic particles tends to be forgotten (Rozumko 2013). In Polish, there are lexemes that can be seen as functional equivalents of German particles (cf. Szulc-Brzozowska 2010), yet the Polish category of particles lacks the formal rigidity of the one in Germanic languages. 3 Based on these facts, one can expect that Norwegian L2 writers with Polish and English L1 background use modal particles less native-like than writers with German L1 background do. 1 It is worth mentioning that Foolen's definition regards pragmatic particles, rather than modal. For the discussion of the various terms, see section 2, cf. also Lind (1996: 177). 2 Since modal particles are often difficult to translate to other languages, we will not provide their English equivalents in the running text. The meaning and functions of those two particles in Norwegian are discussed in section 3. 3 For instance, it is not uncommon to find adverbs or verb phrases labelled as partykuła 'particle' in Polish grammar descriptions (cf. Bogusławski 2003;Kisiel 2008). This may be connected to the fact that in the structuralist view, the term particle is in the broad sense used to refer to any non-declinable part of speech (Möllering 2001: 130-1). Transfer does not need to be the only reason for the more limited use of modal particles in L2 produced texts. It has been shown that certain sequences of words constitute prefabricated sequences which are "stored and retrieved whole from memory at the time of use, rather than being subject to generation or analysis by the language grammar" (Wray 2005: 9). Such formulaic language is the pervasive feature of adult L1 competence, and for language learners it is often considered "the final, and most difficult, stumbling block" in achieving native-like proficiency (Wray 1999: 213). The language of L2 learners has been characterised as formula-light (Ellis 2012). It is thus vital to investigate whether the studied lexemes appear in any fixed, formulaic patterns, in the target language. Adopting this view, we analyse the patterns of use of modal particles in L1 texts, and compare them to L2 texts, to investigate if there are qualitative differences in the way the particles are used by learners and by native speakers.
Finally, the lower frequency of particles in L2 texts may be grounded in the learners' stylistic choices. L2 writers may simply not see the need to use the particle to the same extent as native writers do. For instance, Lorenz (1998) has shown that German speakers of L2 English have a tendency to overuse adjective intensification, which appears stylistically out of place. He speculates that the reason for this tendency may be "insecurity among non-native speakers regarding the effectiveness of their own writing. (…) they might feel a greater need than native speakers to stress the importance -and the relevance -of what they have to say" (ibid.: 59). Another explanation may lie in cultural differences. Petch-Tyson (1998) suggests that writing cultures differ with regard to acceptability of interpersonal involvement in particular discourse types. In her study of four learner groups' writing in L2 English, she has found "that the learner writers are much more overtly present within the discourse than the NS [L1] writers and that this presence manifests itself in different ways" (ibid.: 117). One of the ways can be through statements concerning one's beliefs about the world, also called epistemic 4 stance, which in Norwegian can be expressed by means of modal particles, among others (see section 3).
In the following section, we discuss the status of modal particles within the language system, and present relevant literature on modal particles in L2 use.

Modal particles in a second language
There is a number of terms which can be used to refer to the lexical items we set out to study; in this paper, we choose to use the term 'modal particle' (henceforth MP) 5 . Modality is often seen as a continuum between a narrow understanding, which includes only expressions of necessity and possibility, and the broader understanding as an indication 4 We understand the term epistemic as concerned with the speaker's attitude to the truth-value of the proposition (Palmer 2001: 8).

5
In the literature on the subject, there is an ongoing discussion whether modal particles and discourse markers should be treated as distinct categories or two-sub-classes of one category (cf. Degand et al. 2013). Our choice of the term 'modal particle' is grounded in Norwegian terminology (e.g. Lind 1996 andFretheim 2014), but also in the studies of Traugott (2007) and Fischer (2007). of speaker's stance towards the content of the proposition (Boye 2013: 1). It is the latter understanding that we adopt in this paper. On the grammatical level, the central notion within the domain of modality is the verbal category of mood, while on the lexical level one focuses on "certain auxiliary verbs like must, certain adverbs like maybe, and certain adjectives like possible" (Portner 2009: 1-2). Yet, in studying the semantics of modality, there is a tendency to devote primary attention to modal auxiliaries (Nuyts 2016: 39;Von Fintel 2006), at the expense of other word classes such as adverbs, particles and adjectives.
Studies of modality in a second language have so far suffered from the same bias as the study of modality, with the main focus being on grammatical mood and modal auxiliaries (e.g. Ayoun 2013). Hoye (1997) shares this focus, yet presents also an interesting view on the co-occurrence of modal verbs and adverbs in English, investigating the strength of different collocational links. His study includes an elicitation task, in which he found that L2 English users with L1 Spanish "consistently neglected the potential for modals and adverbs to combine, that is, their collocability with certain types of adverbial" (ibid.: 251). Hyland -Milton (1997) compare English L1 and L2 essays with regard to the use of different lexical items expressing epistemic meaning, and find that there is a discrepancy in how native-like their use in L2 texts is. While the use of lexical verbs, adjectives and nouns is similar for L1 and L2 writers, there are "marked differences in adverbs and modal verbs" (ibid.: 189). The L2 writers seem to rely heavily on modal verbs, whereas L1 writers make use of a greater range and frequency of adverbials. The findings of this study suggest that the adverbial epistemic modifiers are a problematic phenomenon for L2 writers; at the same time, they seem to be vital for how native-like the produced text is perceived (over 41% of all epistemic adverbs in the corpus appear in the essays graded as A or B). A tendency to overuse modal auxiliaries by L2 English writers with L1 Swedish was also found by Aijmer (2002). Interestingly, her study revealed that the overuse stretches also to modal adverbials and modal combinations (modal verb + adverb). Aijmer argues that one possible explanation for this fact is the influence of L1: "Whereas in English, modal meanings are prototypically and frequently expressed by modals, in Swedish, the epistemic modal meaning is more often realised as an adverb or an adverb plus a modal verb. Transfer from the mother tongue may thus reinforce a tendency which is already present in interlanguage to use modal support or to express modality pleonastically" (ibid.: 72).
Aijmer's study shows that L1 modal patterns may be subject to transfer into the L2 performance. Möllering (2001) investigates German MPs, with the assumption that the main challenge that these items pose for L2 learners, is their multifunctionality. Thus, investigating the co-occurrence patterns may serve the purpose of identifying and separating the functions from one another. Using eben as a case study, she finds that the particle frequently appears with existential clauses (that is where the main verb is sein 'to be'), especially in the present tense, and collocates with the modal verb müssen 'must', where the particle "lends emphasis to the obligation of carrying out a particular act. In these clauses, eben serves to express the unavoidability of the obligation" (ibid.: 143). As such, the study lends credit to the idea that in order to study the use of MPs in depth, one should analyse the co-occurrence of those items within the broader scope of a sentence.
There are to our knowledge three studies that shed some light on the use of MPs in L2 Norwegian. In a general study of modal adverbs 6 used by L2 writers, Svensson (2012) found that the particles jo and nok appear early in language acquisition, but do not reach the native-like levels when it comes to frequency and co-occurrence patterns. In spoken L2 data, Horbowicz (2017) found striking discrepancies in the use of modal adverbs and particles: while the range and relative frequency of adverbs was similar in L1 and L2 data, the occurrences of particles were scarce in L2 data and highly frequent in L1 data. In written L2 use, Horbowicz -Janik (2018) investigated the use of MPs by three learner groups (Polish, German and English). Their study showed that the L1 Norwegian speakers used more modal particles than the English, German and Polish learners of Norwegian, and that the difference was statistically significant. Furthermore, the L1 German speakers used the Norwegian modal particles statistically more often (and therefore more Norwegian-like) than the two remaining L1 groups. The fact that there is a significant difference between German learner group on the one hand, and Polish and English learner group on the other hand, might serve as an indication that L1 indeed does influence the choice of modal structures applied in L2 Norwegian texts. However, in order to verify transfer, one needs to adopt a principled method, including both quantitative and qualitative analyses, including the learners' patterns of use of both L1 and L2 (Jarvis 2000(Jarvis , 2010. In the present paper, we do not perform a full transfer analysis, but we use contrastive studies to investigate whether the studied particles have a frequent functional equivalent in translation into learners' L1s (see section 3).

The Norwegian particles jo and nok and their equivalents in Polish, English and German
Although there has been considerable interest in modal particles (or discourse markers broadly speaking) in Norwegian, neither jo nor nok have received as much attention as other MPs. In this part, we briefly discuss jo and nok and their potential equivalents in Polish, English and German. The studies of Polish and English as contrasted with Norwegian are based on translation studies, while for German there are no such studies, so our analysis of German equivalents of jo and nok is mainly based on available literature on German MPs. Solberg (1990: 90) states that jo has a very clear function of expressing that the proposition accompanied by the particle is known or obvious. In other words, jo can be seen as a marker of common ground and possibly also as a marker of epistemic conviction (derived from the shared knowledge). Andvik (1992: 85) suggests, however, that the particle not only marks the information inherent in the proposition as known, but serves a larger interactional goal of opposing "some idea assumed to be 'in the air' (…) by means of recourse to an implied consensus to the facts". In her study of spoken language, Lind (1996: 185) refers to jo as a politeness marker, the use of which is supposed either to signal or to create intimacy between interlocutors. The particle may appear "together with information that must be seen as new for the recipient" (ibid.: 184, our translation). Lind interprets such use as a politeness strategy of preserving the recipient's positive face by referring to shared knowledge, even if this sharedness is only presupposed. Such use of the particle may also be interpreted as manipulative, as the speaker' goal in establishing common ground is to convince the recipient to accept the point of view of the speaker. The most recent study of the item within relevance-theoretic framework suggests that there are two procedural constraints on interpreting the function of the sentence internal jo: to interpret the proposition modified by jo as mutually manifest to speaker and hearer, and to treat the proposition as a premise for deriving and supporting an available conclusion (Berthelin -Borthen 2019: 18). Contrastive translation studies suggest that jo often disappears in translation into English and Polish (Aijmer 1996;Cieślawska 2016;Szymańska 2010). Otherwise, jo is quite consistently rendered in Polish translation by przecież (64% of all rendered occurrences), with other lexemes appearing sporadically (Szymańska 2010). In English however, there seems to be no such consistency: Aijmer (1996) finds that of course and after all prevail in the translation of Swedish ju into English. In German, there are two MPs that can be seen as equivalents to the Norwegian jo: doch and ja. Doch has a clear oppositional meaning (Bross 2012: 191), while ja indicates an affirmative relation between the pragmatic context and the situation expressed in the utterance (Diewald 2007: 132). Bross (2012: 192) claims the common interpretation is that "unstressed ja is typically used when the speaker wants to indicate that the proposition is, should be or can be evident for the hearer". Summing up, we can state that even though Polish lacks the formal category of MPs, there is a single item that corresponds to the Norwegian jo, the lexeme przecież. One cannot name such dominating equivalent for English, where jo is rendered by means of different structures, while in German there are two lexemes (doch and ja) that seem to perform the functions of this Norwegian MP in different contexts. Solberg (1990: 50) suggests that the particle nok is used to express low degree of probability and to weaken subjective assumptions. In Mac Donald's view, nok expresses "some degree of epistemic uncertainty" (1990: 69, our translation), and in some contexts signals negative attitude towards the proposition's truth value. It can thus be considered to convey primarily epistemic meaning.

Nok
Not unlike jo, the particle nok often has zero correspondence in translation to Polish and English (Szymańska 2010). There is no dominating equivalent of nok in Polish, and the range spans from lexemes signalling epistemic certainty (na pewno 'for sure') to those expressing uncertainty (chyba 'maybe', ibid.). For translation into English, Johansson and Løken found that the main group of corresponding lexemes are adverbs, representing nearly 50% of all instances, and "by far the most frequent of these is probably, which represents one fourth of all the correspondences of nok" (1997: 155). These results are in line with Szymańska (2010), and Aijmer (1996) for Swedish. Interestingly enough, Johansson -Løken (1997) found that when nok appeared together with a modal auxiliary, the translation preserved only the auxiliary. This finding suggests that in Norwegian, like in Swedish (Aijmer 1999(Aijmer , 2002, the modal verbs used epistemically need a reinforcement in the form of an adverbial, whereas in English, the modal verb alone expresses the epistemic meaning. In German, one finds two MPs that seem to perform similar functions to the Norwegian nok: wohl and schon. It is mainly the former lexeme that is used to express the speaker's uncertainty (Bross 2012: 194); yet, in contexts when nok accompanies subjective assumptions (e.g. jeg tror nok 'I would think'), schon is also frequent (ich glaube schon). Summing up, the Norwegian MP nok does not seem to have a single corresponding item in any of the studied languages. In Polish, there is a range of adverbs or adverbial phrases that are used as equivalents of nok. In English, there is a tendency to translate nok as adverbs, especially probably, while in German, there is a partial overlap of two MPs that can be used as equivalents of nok.

Data and methodology
The data used for this study comes from a second language corpus ASK, designed and maintained by the University of Bergen. It is a written language corpus of texts produced in an exam situation by L2 Norwegian writers from 10 different L1 backgrounds. Each group is represented by approximately 200 texts, divided equally between two levels of examinations: Språkprøven and Test i norsk -høyere nivå, which roughly correspond to B1 and B2-level, respectively, in Common European Framework of Reference for Languages (Tenfjord et al. 2009). 7 In addition, the corpus contains 200 texts (100 for each test type) written by L1 speakers of Norwegian, which constitutes a control group.
The number of all occurrences of MPs in the three L2 groups and L1 control group is presented in table 1. The data shows that the L1 users of Norwegian employ all MPs more frequently than L2 users, with jo and nok constituting more than 65% of all MPs.
Among the L2 writers, however, there seems to exist a preference for using jo, while nok appears considerably less often. It is worth noting that the numbers in the table represent the number of occurrences, not the number of texts in which the particles appear. Some writers do not use any particles, while others use the studied particles more than once (see for instance example 15 in section 5.4). In the Norwegian control group, if a given writer used any particle, the average number of those was nearly 2 per text, while the same number was between 1,2 and 1,5 per text for L2 produced texts. 8 In the following analysis, we will show the patterns of using jo and nok in Norwegian L1 texts and investigate whether those patterns are implemented in L2 texts written by Polish, English and German learners. Drawing on Möllering (2001), we focus on the following elements of the sentence modified by the particle in question: 1. verb type, where we distinguish between existential verbs (vaere 'be', bli 'become' and finnes 'exist'), modal verbs and other verbs; 2. subject type, where we distinguish between different pronominal subjects (det 'it' 9 , personal pronouns, impersonal pronouns man/en 'one'), generic subjects referring to unspecific groups of people (e.g. mange 'many', alle 'everybody') and other subject types (e.g. nominal phrase, clause, infinitive and others); 3. the co-occurrence of verb types and subject types together with the MP; 4. functional environment, especially co-occurrence with verbs of epistemic meaning, epistemic adjectives, other adverbials or in conditional clauses. 8 The exact numbers for jo: in the Polish L1 group, the 37 particles have been produced by 24 writers (average 1,54). The corresponding numbers are 1,25 for the English L1 group (20 particles/16 writers); 1,31 for the German L1 group (54/41) and 1,78 for the Norwegian control group (75/42). As far as nok is concerned, the numbers for the Polish L1 group is 1,5 (12/8), for the English L1 group 1,27 (14/11), for the German L1 group 1,2 (12/10) and for the Norwegian control group 1,92 (75/39).

9
Det 'it' is a third-person pronoun in Norwegian, so it can be used anaphorically to refer to a previously mentioned referent. However, it is often used as a formal subject without a referent in the previous discourse. In our study, it was impossible to discern whether the use of det was referential or non-referential, so we treat all uses of det as one category.

The particles jo and nok in Norwegian L1 and L2 writingthe patterns of use.
There are several interesting differences in the way L1 and L2 writers use jo and nok. Apart from the quantitative differences (Horbowicz -Janik 2018), there are certain patterns that emerge when looking at the data in more detail.

Verb type
The first criterion of analysis was co-occurrence with certain verb types. In case of both particles, the distribution among the different verb types is nearly equal for L1 writers, while L2 writers show clearer preferences for the existential constructions, mainly with the verb vaere, at the expense of modal verbs, as   Table 3. Number/percent of co-occurrences of modal particle nok with verb types.
There are also differences among the learner groups. The English L1 group uses very few, if indeed any (as in case of nok), particles in connection with modal verbs. The Polish and German groups, however, do use both jo and nok with modal verbs like L1 writers do, albeit with a lower frequency. As far as jo is concerned, the German learner group seems to produce most native-like patterns. As the data of L2 use of nok is rather scarce, one needs to treat the percentage with some caution. We conclude therefore that the German learners follow the use found in the target language to a greater extent than the Polish learner group, which in turn performs more native-like than the English learner group. The modal verbs most frequently used by L1 writers together with jo are kunne and måtte, both mainly in present tense, and the same tendency can be observed for German writers in L2 Norwegian. For nok, the modal verb of by far the highest frequency in L1 writing is ville, used in both present and past tense.

Subject type
While for verb types the tendencies were similar for both particles, there are striking differences in how the writers use jo and nok with different types of subjects. As far as jo is concerned, the numbers presented in table 4 show a great variety of uses for all four studied groups. The patterns are roughly the same, with all groups using mainly pronominal det and the impersonal pronoun man/en. In the English L1 group, one can see a certain discrepancy, as the particle is often used with the first-person pronoun jeg, which in the other three groups is rather uncommon.  Table 4. Number/percent of co-occurrences of modal particle jo with subject types.

JO
In the case of nok, however, the picture is rather different, as can be seen in table 5. Among the learners, there is a visible tendency to use either det or nominal phrases, whereas L1 writers often use the particle together with the first-person pronoun jeg and generic subjects referring to a certain part of population, such as mange 'many (people)'¸ de fleste 'most (people)' and noen 'some (people)'. This is indeed the predominant subject type found in L1 writing, except for other subjects (which in most cases includes nominal phrases and clauses).  Table 5. Number/percent of co-occurrences of modal particle nok with subject types.

'Subject + verb + particle' patterns
If we combine the data from the two previous sections, one can notice certain patterns that are present in L1 writing and lack (except for one case) in L2 writing. In case of jo, the only pattern that is found in texts produced by all four groups is the structure: det er jo, as exemplified by following sentences: (1) Det er jo umulig å tvinge noen til å lese. (PL/h0479) 10 It is jo impossible to force someone to read.
(2) Det er jo koselig med bekjente og de fyller en stor rolle i livet. (EN/h0470) It is jo nice to have acquaintances and they fill an important role in life.
(3) Det er jo sant at SMS er ikke noe tegn for vennskap (…) (GE/h0432) It is jo true that text messages are no sign of friendship (…) (4) Det er jo slik for endel type transplantasjoner at (…) (NO/h2021) It is jo like that for a number of transplantation types that (…) The L1 control group repeatedly makes use of the patterns including modal verbs and either det or man as subject: (5) (…), kan det jo tyde på at ALT dreier seg om å vinne. (NO/h2071) (…), can jo indicate that everything is about winning. (6) Man vil jo så klart svare nei på dette (…) (NO/h2060) One would jo clearly want to answer no to that (…) Those patterns are also visible in the texts produced by the German learner group (4 out of 11 modal verbs used with the particle jo appear with det as subject), while they are not observed in the Polish and English learner group.
As far as nok is concerned, the pattern that predominates in L1 writing is the use of a generic subject accompanied by vil/ville and the particle nok (example 7 and 8).
10 The codes for examples give the first language (PL -Polish, EN-English, GE -German, NO -Norwegian) and the author's ID-number from the ASK-corpus (numbers beginning with "s" stand for Språkprøven, while the ones beginning with "h" represent Bergenstesten). For the sake of clarity, we use dynamic translations, instead of translating word by word, while the particle is left untranslated. The examples have been rendered as they are, so they can include various deviations from the Norwegian language norm (also in case of Norwegian L1 texts).
(7) Mange vil nok tenke på at en annen har dødd" for" å gi bort et organ (…) (NO/h2093) Many will nok think that someone else has died to give away an organ (…) (8) Noen, eller kanskje de fleste ville nok si nei til dette. (NO/s2079) Some, or maybe the majority would nok say no to that. This pattern is not to be found in any of the three learner groups. The only pattern with nok that is frequent in L2 produced texts is the particle used with the pronoun det and the verb vaere 'to be', that is det er nok. 11 The L1 writers do indeed use the structure, but again, in many cases the pronoun det is only the formal subject, while the logical subject is again a generic unspecified adjective phrase, as example (9) shows: (9) Her er det nok mange som igjen vil se det som forkastelig (…) (NO/s2035) Here there are nok many who again would see it as condemnable (…)

Functional environment
Many of the propositions which the particle jo is modifying, are grounded in the writers' (presupposedly shared) knowledge of what the world looks like. This is true for the Polish and German learner group, as well as the L1 control group, while in the English learner group, such usage has not been observed, possibly also due to the fact that in this group, jo often modifies sentences with first person pronoun as a subject.
It can also be combined with the modal adverb selvfølgelig or selvsagt (which correspond to 'of course' or 'self-evident' in English), or with modal adverbs showing subjective judgement of the proposition's desirability, like dessverre 'unfortunately' or heldigvis 'luckily'.
Yet, there is one usage that none of the studied L2 groups employs, and which could be labelled as concessive. It is when the proposition modified by the particle acts as a hedge on what has been said previously, as in the following examples.
(15) Man kan jo spørre seg om resten da er tapere? Man vil jo så klart svare nei på dette, og i en viss forstand har man jo rett. (NO/h2060) One can jo ask oneself if the rest then are losers? One would jo clearly say no to that, and in one way one is jo right. (16) Det kan jo selvfølgelig diskuteres. (NO/h2071) It can jo be discussed.
It performs a vital role of presenting the opposite opinion as equally acceptable as the writer's views, thus showing a greater degree of detachment and lowering writer visibility (terms after Petch-Tyson 1998).
The particle nok as used by L1 writers often appears in hypothetical constructions, either full conditional sentences, as example 17 shows, or only parts of those, as in 18. The particle nok can collocate with a modal adverb of similar epistemic strength, kanskje 'maybe', as we can see in example (8). Such use does not appear in learner texts, but the frequency of such collocation in the L1 corpus is too low to draw far-fetched conclusions. Horbowicz -Janik (2018) have shown that L2 Norwegian writers use fewer modal particles than L1 writers, and that there are statistically significant differences in the frequencies of using the class of MPs among the studied learner groups. In this paper, we have focused on a qualitative analysis of the data, contrasting the patterns of use that predominate in L1 and L2 produced texts. In the following section, we will address the two issues that are the scope of this paper, namely the characteristics of the use of jo and nok in Norwegian L1 and L2 writing, and the account of differences among the three learner groups.

Jo and nok in Norwegian L2 writing contrasted with L1 writing
The analysis has shown that the L2 writers are more restricted than L1 writers in their range of patterns employing jo and nok in their texts. One main finding is that none of the learner groups make use of the collocation modal verb-modal particle to the extent the L1 writers do. Several studies have shown that L2 writers tend to overuse modal auxiliaries at the expense of other modal structures (e.g. Hyland -Milton 1997;Aijmer 2002). In his study of L2 texts collected in ASK as compared to L1 control group, Svensson (2012: 172-79) found that modal verbs differ when it comes to their frequencysome of them are used more often in L2 texts than in L1 texts (e.g. skal and må), while others appear less often (vil). In this study, we have not investigated the frequency of modal verbs, yet it would be reasonable to check whether L2 writers' failure to produce modal verb-modal particle combinations does not arise from the lower frequency of specific modal verbs, or modal verbs in general.
The fact that the L2 users do not employ modal verb-modal adverb collocations in their writing, is in line with the claim put forward by Hoye (1997) that L2 speakers of English tend to neglect the possibility of combining modal verbs with adverbs. At the same time, it does not support the claim offered by Aijmer (2002: 72) that non-native writers tend to express modality pleonastically. It is worth mentioning that Aijmer (ibid.) investigated users of English with Swedish L1. As already stated, the Swedish modal verbs are less lexicalised in their epistemic meaning than their English counterparts, so the epistemic reading needs to be marked by means of collocating with a modal adverbial. Thus, the findings of Aijmer's study could probably be interpreted as a case of transfer from L1 to L2. As shown in table 2, the German learner group uses the particle jo considerably more often with modal verbs, and thus in a more native-like way than the Polish and English learner group. The same cannot be said for the particle nok, yet the limited number of occurrences in all three L2 groups does not allow further generalisations.
As far as the realisations of L1 patterns for both particles are concerned, the L2 writers seem to prefer the existential verbs, especially vaere together with the third-person pronoun det. This finding may suggest that this use is seen by L2 learners as more prototypical in the target language, and is as such acquired quicker than other, less prototypical uses (cf. Kellerman 1979). One could also argue that such constructions are retrieved from the learners' memory as a prefabricated pattern rather than actively generated in the process of language production (Wray 2005: 9). This hypothesis requires corroboration in data from bigger language corpora, and possibly from some experimental elicitation tasks. As Norwegian modal particles are typical for the spoken language (Lind 1994), it would be worth checking whether the construction det er MP (jo/nok) is especially frequent in this medium.
Why do L2 Norwegian writers not employ the particle nok to the same extent as they use jo? There may be several reasons for this fact. One of them may be negative transfer resulting in underproduction. As discussed in section 3.2, nok does not have a good functional equivalent in translations into Polish and English. Yet in German, the situation is different: there are two lexemes that in different contexts can be seen as equivalent to nok, and both are classified as MPs. This leads us to believe that there are other reasons for the particle's underproduction in L2 writing. One of them may be that that the particle does not appear in a specific pattern with the frequency that allows for the acquisition of the pattern as a formula (Wray 2005). As tables 4 and 5 show, there is a clearer division of subjects that collocate with the particle jo than with the particle nok. On the other hand, both jo and nok seem to follow some predominant patterns. In case of nok, there are two important patterns visible in the L1 corpus. One of them is the extensive collocation with the modal verb ville, and the other is the use with unspecific generic subjects like mange 'many (people)' or noen 'some (people)'. Svensson (2012) has indeed found that the modal verb ville appears with a lower frequency in Norwegian L2 writing than in L1 writing, yet we do not know if this holds for the three L1 groups studied here. 12 The fact that those existing patterns are not found in L2 writing, may suggest a different explanation for the lack of their implementation, namely a pragmatic reason. L2 writers may simply not see the need to use the particle to the same extent as native writers do, as shown by Lorenz (1998) in connection to adjective intensification. It is possible that the particle nok¸ signalling weak epistemic conviction and thus acting as a hedge on what the writer is claiming, is perceived by the L2 writers as undesirable, as it weakens their stance. One can also ground their choices in cultural differences in the writing cultures, as posited by Petch-Tyson (1998). The reason for learners' frequent use of jo as opposed to nok can be connected to their wish to manifest their voice in the text. Since jo serves to reinforce the views presented by the writer as grounded in common knowledge, it makes the writer's presence more overt than in the case of nok which weakens the epistemic value of the statements.

Differences among L2 groups
As already stated, there are several differences between the investigated groups of learners. In the case of jo, the differences are both quantitative and qualitative, while in the case of nok, only the latter can be observed. The English learners differ the most from the L1 control group as they fail to use particles with modal verbs, show a different pattern for the choice of subjects and do not employ particles in a similar functional environment (expressing beliefs about the world in case of jo, and hypothetical constructions in case of nok). The Polish learners use the studied particles with modal verbs to some extent, yet not with the same frequency as L1 writers. At the same time, they seem to use the particles in the same functional environments as the control group. The German learners come closest to following the patterns found in L1 writing.
How can the differences between the three groups of L2 writers be accounted for? It seems that the German learners profit from having the category of modal particles in their L1. The German MPs can be used with modal verbs (cf. Möllering 2001), a pattern that they can successfully "copy" into their L2. The differences between L1 and L2 may have a negative effect on the English learner group who are the least native-like in their use of particles. The Polish learner group, on the other hand, is more difficult to discuss. Polish lacks the formally limited category of modal particles, yet exhibits several lexemes that often perform the same function as MPs in Norwegian. The degree of the Polish writers' successfulness in using jo in Norwegian, can thus be attributed to lexical transfer, as we indicated in the contrastive studies (see section 3). In order to confirm this hypothesis, one would need to collect data from Polish L1 and compare those with the Norwegian L2 use of the Polish learner group. The same procedure could be applied to the German use of the particle jo.

Conclusions and implications of the study
In this paper we have shown that Norwegian L1 and L2 writers follow different patterns of using particles jo and nok, and that those patterns also differ among the studied learner groups. We have identified typical syntactic and functional environment for L1 produced particles. As the findings are based on limited number of occurrences (75 in each case), it would be justified to check the L1 patterns on a bigger corpus of data. Yet, ASK-corpus offers a unique possibility to study personal argumentative writing where the use of particles does seem natural and in the right place. The use of newspaper corpora might not yield such promising results. It has to be pointed out that the topic of essay may influence the frequency of particular modals (Hinkel 1995, after Aijmer 2002. Due to space limitations, we were unable to investigate this issue in detail, yet it is worth mentioning that both L1 and L2 writers wrote essays on similar topics; the learner corpus and the control corpus can thus be considered comparable. The implications of the study are two-fold. Firstly, the findings suggest that modality in L2 Norwegian is a promising field of study. As we have seen in the discussion, there can be several reasons for the L2 writers not to use the particles in question: transfer, general interlanguage tendencies or cross-cultural factors. Whatever reason, the consequences are that the L2 produced texts may fail to conform to the L1 expectations, thus resulting in not achieving the intended communicative goal, or purely practically, receiving a lower grade in an exam situation (cf. Hyland -Milton 1997). In studies conducted on ASK-corpus, Golden -Kulbrandstad (2013) and Golden et al. (2017) have investigated why essays written by one L1 group (Vietnamese) have been rated lower in a subsequent re-evaluation than another L1 group (Spanish) who showed a similar degree of formal correctness. One of the variables that came under scrutiny, was "the degree of agency and personal presence negotiated in the texts" (Golden et al. 2017: 256), manifested by use of pronouns and their combinations with certain verbs. Their studies conclude that although there are several variables that lead to a given text's high or low rating, there is also a possibility of cultural transfer of textual norms, which in this case could have influenced the Vietnamese texts to "seem more distant and difficult to interpret" (ibid.: 267). It is reasonable to claim that the modal structures used by L2 writers, which play a crucial role in modifying the propositions in terms of speaker's convictions and involvement of the reader, may also influence the examiner's general perception of the author of the text as more, or less, proficient language user.
As modal particles often are especially frequent in conversational data (Lind 1994), it would be interesting to investigate their collocations and the functions of those in spoken discourse, as for instance Adolphs (2007) does for certain English collocations of modal verbs and modal adverbs. It is also possible that the L2 writing reflects the spoken use of particles to a greater extent that the L1 writing does, due to a greater genre sensitivity of native speakers. The scope of this paper does not allow for further discussion of this question, which would probably require an experimental study to be investigated in detail.
Secondly, the study may have didactic implication for teaching different modal structures, including modal particles, in Norwegian L2. So far, few studies have dealt with the subject, one notable exception being Borthen et al. (2016). Our study has shown that the L2 use of modal structures differs considerably from the patterns found in L1, which may thus influence the perception of how advanced the learners' written language is. One should therefore devote attention to teaching modal patterns, as "it can facilitate the development of both accuracy and fluency" (Hunston 2002: 167). Yet, in order to do that, there is a need of further studies into the use of particles in Norwegian written and spoken language.