Does Focused Teaching of NP Elaboration Enhance Young Learners’ Narrative Writing Competence in English as a Second Language?

In the present investigation, 30 students in grade 4 (aged 10-11) within the Swedish school system were, with intervals of two weeks, asked to compose three narrative texts in L2 English. With the aim of enhancing their narrative writing skills, the learners received explicit teaching of NP (noun phrase) elaboration (both preand post-modification) prior to the first two texts, whereas no such instruction was offered before the third story. Two control groups, one writing in Swedish (L1) and one writing in English (L2), neither being offered treatment before doing so, were also incorporated. Despite some initial problems, the results show that the treatment the learners received enhanced their narrative writing skills in a number of ways. Text 3, in contrast to Text 1, does not only display an increase in text length, but also a growth in the number of post-modifying prepositional phrases and relative clauses (the latter primarily having general discourse functions but also narrative discourse functions), as well as an increase in NP and VP coordination, all being signs of more advanced narrative writing.


Introduction
In a first language, acquiring the skill to write is one of the most basic parts of a child's schooling in today's society. However, while some students may never fully rise to the challenge involved in composing a cohesive, coherent and imaginative narrative, the great mass, based on the present author's own experience, appear to produce texts of average quality, mostly perhaps succeeding in getting the message across but not in very refined ways. So, why is it so difficult to excel at one of the most fundamentally important parts of literacy?
The answer to this question may partly be found in Strömqvist, Nordqvist & Wengelin (2004). The study investigates the L1 development of speaking and writing in three age groups: 9-12-year-olds, 15-year-olds and adults. The findings indicate that while 9-12-year-olds still largely maintain a mindset geared towards spoken language when speaking as well as writing, 15-year-olds have developed a specific mindset for written language. That is, whereas 15-year-olds have realized that there, in most cases, are marked differences between spoken and written language and have developed means to express such differences, 9-12-year-olds have not. This means that while younger age groups still tend to add written information mainly by linking one clause to the next, using primarily the conjunction and (Christie & Derewianka, 2008), older age groups are capable of displaying more sophisticated ways to pack the written message they want to convey, one being the elaboration of noun phrases (NPs) (Svensson, 2018).
The situation grows in complexity when what is being learnt is not a first, but a second language. While parts of learners' new mindset geared towards written language may be automatically transferred to a second language, the fact that, in addition to vocabulary and spelling, new forms for already established functions have to be learned, may be a hindrance to complete transfer. The aim of the present study is therefore to explore to what extent explicit teaching of NP elaboration may help improve young learners' narrative writing in English as a second language.

Theoretical Background 2.1. Young learners' narrative writing competence and development
In the initial stages of writing, there is what is referred to as pretend writing, in which 'texts' consist of various doodles. These are most likely attempts at imitating what children have observed their parents do, and/or induced by a variety of kindergarten activities (Svensson, 2018). Then comes synthesis-in-analysis writing, which displays clusters of primarily consonants without any clear division between words. This stage is followed by limited efficient writing, in which texts slowly start to look like adult writing (e.g. Björk & Liberg, 1999;Donovan & Smolkin, 2006;Kamberelis, 1999). Stories composed at this level often include drawings, to which short verbal texts are added. Once these texts start to describe simple events and reaction to these events (the embryo of problemoriented plots), children have reached what Bamberg & Reilly (1996) refer to as the middle phase (around the age of six). Unlike more advanced writing, texts belonging to this phase still normally adhere to patterns typical of spoken language (as does the choice of vocabulary), where one clause is added to the next, primarily with the help of the conjunction and (but also but and so) (Christie & Derewianka, 2008). These stories usually also include a great deal of repetition of similar events and subjects (Svensson, 2018), overuse of one and the same structure generally being the only type of strategy children have at their disposal to expand a story at this stage. This repetition of specific structures, commonly referred to as syntactic priming, appears to prepare a learner for the processing of other types of sentences that are concerned with the same type of structure (Clealand & Pickering, 2005;Loebell & Bock, 2003), later allowing an exploration of new and more refined ways of conveying different types of information. In the final phase (around the age of nine), children are beginning to offer descriptions of characters that affect major parts of the entire plotline of the story. That is, no longer do these descriptions only create links in the linear organization of the events of a story as described above, but also help create links to parts that are non-consecutive. Children's stories now usually become longer (Bamberg & Reilly, 1996;Nordenfors, 2011), and it is also around this age that children start to incorporate formulaic expressions into their writing (Svensson, 2018).
Furthermore, once more adult-like writing starts to emerge, its development, Berman & Slobin suggest (1994), occurs in form-function relations. More precisely, Berman & Slobin propose that narrative writing development occurs through a reciprocal relation between form and function, so that the acquisition of various forms triggers more elaborate narrative functions, which in turn prompt the need for acquiring new linguistic forms. This assumed dependency is therefore also able to capture how these relations develop over time, e.g. how old forms take on new functions and how new forms are acquired to be able to express new narrative functions as the writer becomes increasingly sophisticated in conveying his/her message. Berman & Slobin discuss five types of formfunction relations that are crucial to the construction of a narrative text and its development. The most important category, from the perspective of the present investigation, is connectivity, which is concerned with how individual sentences become texts that are cohesively and coherently linked. Pre-and/or post-modified NPs and syntactic coordination thereof are examples of forms that may improve connectivity in a text, thus making it more advanced. Moreover, the extent to which these formfunction relations are present in a text, as well as their elaborateness, is dependent on three so-called guiding themes: filtering theme, packaging theme, and developmental theme, the latter two being of main interest to the present study.
The developmental theme acknowledges the fact that young learners generally have fewer options when conveying a message. The main reasons are that, in contrast to older learners, they are limited 1) cognitively, i.e. they are not able to understand different perspectives, 2) communicatively, i.e. they are not able to completely determine the listener's/reader's point of view, and 3) linguistically, i.e. they do not master all the linguistic tools available.
The packaging theme deals with the vertical construction of a story. That is, it is concerned with how events, to begin with usually linked in a linear fashion, become increasingly more elaborate through the acquisition of more sophisticated ways to package the information in a hierarchical structure. Compare, for example, The little girl came home. She went into the kitchen. She drank some juice, which offers equal importance to the events, with After getting home, the little girl drank some juice, where, with the help of subordination, the activities have been packed into one complex unit, also forcing the listener/reader to infer parts of the message (see also Svensson, 2019). The first type of packaging to occur in the writing of young learners is usually temporal packaging, the when-clause (when the girl saw the dog, she ran to it) typically appearing first. Berman & Slobin also mention other types of packaging, the most relevant of which to the present investigation is the relative clause (which will be discussed in more detail in the sections to come).
Lastly, narrative development is not always a linear process in which increasingly more complex structures are added to a learner's repertoire, but usually also includes 'waves of stability and turbulence' (Sensson, 2018: 246) in which, for instance, reuse of old forms are found alongside the use of newly acquired forms (e.g. Evensen, 2005;Ortega & Byrnes, 2008). Such waves of turbulence and stability display similarities with Dynamic Systems Theory (e.g. de Bot, Lowie & Verspoor, 2007;Larsen-Freeman & Cameron, 2008;Meara, 2009;Plaza-Pust, 2008), according to which a linguistic system is made up of interconnected parts where some elements may be more tightly linked than others. This implies that 'a gentle nudge' to one part of a system may very likely create 'ripples' that will lead to changes in the rest of the system. The focus of the present study is to explore what changes in L2 narrative writing development focused NP elaboration might yield. There are, however, other theories that predict that if the difference between the level of knowledge a learner possesses and the level of complexity of the intended instruction is too great, the result may be negative rather than positive. This is, for instance, what is put forth by the 'teachability hypothesis', which forms an important part of Processability Theory (Pienemann, 1998). (See also Krashen (1981) for the same claim.) The hypothesis thus claims that instruction is only beneficial if it focuses on structures belonging to 'the next stage', as no stages of acquisition can be ignored, not even with formal instruction. The same is suggested by the Zone of Proximal Development (Vygotsky, 1978), proximal referring to the skills the learner is close to mastering. Vygotsky here suggests that there is a difference between what a child can achieve cognitively on its own and what s/he can do together with a more experienced individual, for example, a teacher, but also that the teacher should focus on what naturally would come next in the learner's general developmental trajectory. Both these theories hence suggest that focused NP elaboration may not be beneficial to all learners.

The NP in increasingly more complex narrative writing-an SFL approach
One way to describe narrative writing development is from a systemic functional perspective (Bamberg & Reilly, 1996). Within Systemic Functional Linguistics (SFL), the construction of a text simultaneously involves an exchange of content, a negotiation of roles between, for example, writer and reader, and a structuring of text. In other words, an SFL framework, just like the psycholinguistic approach discussed in Subsection 2.1, relates forms to functions, and is, as a learner's mastery develops, also capable of capturing changes in these relations over time.
Putting together a text involves consistently making lexical choices as well as choices regarding the transitivity pattern of a clause. The latter includes the process itself (i.e. the verb phrase (VP)), the participants in the process (subject and object) and the so-called circumstances (i.e. adverbials (AdvP) or prepositional phrases (PP)). Participant roles are realized as noun phrases (NP). As an NP can be elaborated on with preand post-modifiers, enabling writers to condense information in one and the same clause, this phrase type is, according to Halliday, 'a powerful resource for meaning-making ' (1998: 196).
An NP may contain five functional elements, and it is structured as displayed in Table 1: The key element of a noun phrase is the thing (head). This may be preceded by a classifier, which is either an adjective or a noun. Epithets (adjectives), which may or may not be intensified by adverbs, also function as pre-modifying elements of the thing, and as such offer additional descriptions. Deictic elements, as seen in Table 1, precede epithets and classifiers. These are demonstrative, interrogative and possessive pronouns, and nouns in the genitive form. Of the three pre-modifying elements, deictics have the greatest specifying potential whereas classifiers have the least specifying potential. The latter type is instead a permanent attribute of the thing, i.e. the thing and the classifier together form a tightly knit unit (Martin & Rose, 2007: 96). The thing may also be elaborated on further with the inclusion of post-modifying elements, which in SFL are termed qualifiers. These are primarily prepositional phrases (PPs) or relative clauses. Other types of post-modifying elements are when-clauses and infinitive clauses. As for post-modifying relative clauses, Dasinger & Toupin (1994: 462) place different functions of these elements along a developmental continuum (see also Berman & Slobin (1994). The first four functions, the ones that are likely to appear in earlier stages of writing, have general discourse functions, whereas the remaining five functions, appearing in later stages, display specific narrative discourse functions. Put differently, Dasinger & Toupin here make a distinction between relative clauses that may appear in any type of text and those whose functions will not only add to the building of a cohesive and coherent text, but will also help create an imaginative, rather than flat and uninteresting, story. Table 2 illustrates this continuum. In their investigation, Dasinger & Toupin (1994) were able to observe a growth in the use of these relative clauses and their functions in relation to age. Table 2. Dasinger & Toupin's (1994: 462) continuum of increasingly more difficult post-modifying relative clauses.
Relative clauses General discourse functions 1 NAME (naming referents) 2 SIT-NEW (situating new referents) 3 SIT-OLD (situating old referents) 4 REID (reidentifying old referents) Narrative discourse functions 5 PRES (presenting main characters) 6 MOT (motivating/enabling actions) 7 CONT (continuing the narrative) 8 EXP (setting up expectations about narrative entities and events) 9 SUM (summing up past and upcoming events) It should be added that what elements are included in an NP usually varies in accordance with genre. While NPs tend to be very short in casual conversation, consisting of either a pronoun only or an article + head with a possible epithet, scientific texts of various kinds may, in many of the NPs, include most or all of the pre-and post-modifying elements described above, thus making it possible to, for instance, display, quantify, classify and qualify in one and the same phrase. 'In narrative texts', Svensson writes (2018: 28), 'we expect to find a number of describing pre-modifiers (adjectives, often together with adverbs, expressing for example intensity and force), and various post-modifers'. Put differently, the presence of elaborated NPs is a typical characteristic of narrative writing.
Before continuing the discussion of pre-and post-modification of noun heads, it should be mentioned that learners go through several stages regarded as precursors to elaborated NPs. Plunkett & Strömqvist (1992), for instance, mention juxtaposition of head and descriptive word (the ball, it's red) as well as adjective in complement position (the ball is red), which will both eventually lead to pre-and post-modified NPs (the red ball, the ball that's red).) Svensson (2018), while providing no quantitative insights into NP elaboration over time, offers a thorough picture of qualitative development. She follows the writing development, divided into five increasingly more complex phases, of one L1 learner of Swedish from her first year in school (age 7) to school year 12 (age 18), and more. As phases 4-5 include texts that are far beyond what the informants in the present investigation are capable of producing, what follows will only include NP elaboration as described in phases 1-3.
In the first phase, when the informant is between seven and nine years old, the first elaborated NPs start to appear. In the first story, written at the age of seven, only one single elaborated NP (epithet + thing) is included. Text 2, written the following school year, incorporates two elaborated NPs, both of which again include pre-modification only. The last of the stories and also the longest, written at the age of nine, includes not only a more diverse pattern of elaboration, but also the first post-modifying elements, namely three short relative clauses. While two of these have general discourse functions (SIT-NEW and NAME), the third one displays a more complex function specifically aimed at developing the narrative (EXP). Svensson comments that at this early stage of writing, the pre-and post-modified head is the single most important tool available to her informant to condense information (2018: 68).
In school years 4-6 (ages 10-12), the learner's use of elaborated NPs increases and diversifies even further. For example, in Text 4, written at age 10, the first post-modifying PP appears (Elsa with her smile). Additionally, more and more NPs become simultaneously pre-and postmodified. For the first time, Svensson also mentions the existence of coordinated noun phrases (e.g. Jonathan and his girlfriend Sarah), even though not all of them are necessarily elaborated. Some coordinated VPs (stood there and talked) are also found, though they are fewer in number than the coordinated NPs.
Furthermore, a total of 19 relative clauses, restrictive as well as nonrestrictive, referring to people, things as well as places, occur in these texts. While 13 still have general discourse functions (SIT-OLD and REID being new ones), as many as six now display narrative discourse functions (PRES and MOT being new ones). This is thus one example of how old forms are used for new functions as the learner becomes more advanced. At this stage, these post-modifying elements, Svensson emphasizes (2018: 98), constitute the most important feature in the tools available to package a great deal of information in one and the same clause.
In Texts 12-14 (ages 13-14), finally, both pre-and post-modifiers develop dramatically in complexity. The most conspicuous feature in these stories, Svensson notices (2018: 128), is, however, the increased use of classifiers, which results in a more topic-specific vocabulary. Also, coordination of elaborated noun phrases (e.g. dark green stripes and pink dots) now seems to be a relatively common feature. Moreover, these texts include nine relative clauses, two of which display narrative discourse functions not seen before (CONT and SUM). During this phase, Svensson's informant hence continues to develop new functions of already established forms.

Differences in the NP between Swedish (here L1) and English (here L2)
English as well as Swedish adheres to the NP structure presented in Subsection 2.2. There are, however, some differences in form that need to be pointed out.
Firstly, while Swedish indicates gender and number morphologically in pre-modifying adjectives, English does not. Moreover, the use of gender and number is here completely arbitrary, as is also the case with the (in)definite articles, although once a decision has been made there is alignment among article, gender and number. ett stort hus-det stor-a hus-et-de stor-a hus-en (a big house-the big house-the big houses) en stor cykel-den stor-a cykel-n-de stor-a cykl-arna (a big bike-the big bike-the big bikes) In these respects, native Swedes who are learning to write in English therefore make a transition from a rather complex system to a considerably less convoluted one, which may of course help promote their NP elaboration in English as a second language.
When it comes to post-modifying relative clauses, Swedish learners instead encounter a considerably more complex system in English than what is present in their own language. While English distinguishes between who, whom, which and that, Swedish may use som. This means that whereas the Swedish relative pronoun som is a highly accessible form (Dasinger & Toupin, 1994), which in turn may increase its use and consequently accelerate its mastery, this is not the case in English, where the diversity of relative pronouns may instead constitute an obstacle to the use of relative clauses by Swedish learners. (Even though errors in form are ignored in the present study, the point here is that the mere fact that learners are faced with quite a few alternative forms, may (subconsciously) deter them from incorporating elements from such a system in their texts.) It should here also be pointed out that, similar to English, Swedish possesses the possibility of excluding the relative pronoun when used in object position in a restrictive relative clause. (For more on the use of relative pronouns in Swedish, see Plunkett & Strömqvist (1992).)

The Present Study 3.1. Research question addressed
One research question is addressed in the present investigation: Does explicit instruction of NP elaboration enhance young learners' narrative writing competence in a second language?

Informants and method
Based on availability, a total number of 50 learners (from the same school, and native speakers of Swedish and Swedish only) took part in the present investigation, all of whom were middle school students in grade four (ages 10-11) at the time of the collection of the data. There are two main reasons why this age group was selected to be included in the study. Firstly, within the Swedish school system students most commonly do not start to learn English until the third grade (age 9). Thus in order for them to acquire enough vocabulary to be able to produce running text, i.e. sentences displaying textual cohesion as well as narrative coherence, it seemed, after going through a great deal of material, necessary to wait at least a year (see also Keaveney & Lundberg, 2014). Secondly, as the NP appears to grow considerably in complexity (pre-and post-modification) in Swedish as a first language during these years (Svensson, 2018), it was believed that focused teaching in English would be able to set the wheels in motion in a second language too. 30 of these 50 informants made up the experimental group. They were asked to write three narrative texts each. Prior to writing the first essay, the students received explicit teaching of NP elaboration. The sentence The dog is sitting on the floor was written on the whiteboard and the learners were asked to focus on the head of the subject noun phrase (dog). Several examples were given of pre-modification (deictics, epithets and classifiers) as well as post-modification (prepositional phrases and relative clauses). In the class as a whole, the students were then asked to come up with their own examples. (It was noticed that while pre-modifiers were offered quite spontaneously by the learners, the students had to be prompted to produce post-modifiers to a considerably greater extent.) Next they were asked to write a short story in English about the dog.
Two weeks later, the same 30 learners were asked to write a second composition, before which they again received treatment. This time the phrase My best friend was written on the whiteboard, and the students were, in the same way as described above, again offered examples of preand post-modification. Similar to the first treatment session, the learners were then requested to give some examples of their own, after which they were again asked to write a composition in English on the topic given. (As in connection with the first treatment session, pre-modifiers seemed more easily offered than post-modifiers.) Still another two weeks later, the same 30 students were asked to write their third and final essay. An empty shop window was drawn on the whiteboard. It was put to the students that they could write about what could be found in the window, what they would like to buy for themselves or for someone else and why, but it was emphasized that they were allowed to write any story as long as it had something to do with the shop window. Most importantly, this time no focused instruction regarding NP elaboration preceded the writing of the essay.
Regrettably, due to a misunderstanding regarding scheduling no pretreatment data were collected with these 30 students.
The remaining 20 testees constituted two control groups, in which neither case NP elaboration treatment was offered prior to the composition writing. These students were also asked to write about the shop window, 11 writing in L2 English and nine writing in their native language (Swedish), the former group thus replacing the missed pre-treatment opportunity with the 30 students discussed above. (It was due to the many similarities in NP structure between English and Swedish (see Subsection 2.2) that the present author decided to ask one of the two control groups to write a narrative in their mother tongue (Swedish).) In none of the cases described above, experimental group or control groups, were there any time constraints.
Lastly, while the 30 students in the treatment group and the 11 students in the control group composing in English wrote their essays during the spring term (beginning of March) of their fourth school year, the nine students in the control group composing in Swedish wrote theirs in the autumn term (just before Christmas) of the same school year, i.e. around two and a half months before the other students.
What has been explained above, is summarized in Table 3. The table also provides information about the total number of words in the compositions, the average number of words as well as the lowest and highest number of individual words. As the reader can see, whereas the English essays written after treatment display great differences between individual performances in all three essays, there is a steady increase in the total number of words and average number of words from Essay 1 to Essay 3. In addition to the above, the total number of nouns, verbs, adjectives and adverbs were counted in the different treatment/control groups. The relevance of this will become clear to the reader in the result and discussion section.

The categorization and counting of the NPs used in the students' essays
Except for single pronouns, all noun phrases, even those within other phrases and clauses were counted. In an initial step, the NPs found were divided into those that were elaborated and those that were not. Here errors with articles (e.g. *an tiny dog) and discord in number (e.g. *three tiny dog) were ignored. Elaborated NPs were then subdivided into those that were used and/or elaborated correctly and those that were not. The unnatural and incorrect uses will be explained in more detail in the result and discussion section.
Next the elaborated NPs (correctly and naturally used) were subdivided further. Here three additional categories were formed: NPs that were pre-modified, NPs that incorporated post-modifying elements and those NPs that contained both types of modification. In accordance with an SFL framework, as explained in Subsection 2.2, deictics, epithets (including adverbial intensifiers) and classifiers were considered premodifying elements. Within the group of qualifiers, in accordance with Subsection 2.2, prepositional phrases, relative clauses and a group of miscellaneous post-modifiers were considered.
Lastly, NP density was calculated by dividing the number of noun phrases by the number of clauses, an increase in density generally being considered as indicative of more complex written language.

Teacher assessment
In a final step, without being given any information about the focused NP elaboration instruction, an experienced middle school instructor, having taught Swedish as an L1 and English as an L2 for more than 20 years, was asked to assess the achievements of the 30 students in the treatment group. While Swedish students in the fourth year are generally not officially given grades, teachers often make use of unofficial descriptions of students' levels of performance. This particular middle school teacher described how she usually implements a three-point scale in connection with all types of writing tasks: 'extra support needed', 'knowledge acceptable' and 'knowledge more than acceptable'. This is also the system she used in connection with the present investigation. Here 'extra support needed' (ESN) includes the following criteria: Too little text to make an assessment. The content is not comprehensible. Rules of writing are implemented with considerable uncertainty and there are deficiencies in vocabulary use. Uncertainty of language use. Strategies to make communication purposes clear are missing.
'Knowledge acceptable' (KA) is defined as follows: Vocabulary use shows certain variation. The student displays a good attempt at communicating written context. The text is quite simple, yet it is comprehensible in terms of phrases and sentences. Spelling errors/mistakes exist, but do not disrupt communication. There is grammatical uncertainty, but the student still succeeds in making himself/herself understood. The student is able to make use of strategies that solve problems and thus improve communication.
'Knowledge more than acceptable' (KMTA) is described in the following way: Vocabulary use is varied and enhances the quality of the text. The text is clear and relevant in relation to the task given. Sentence structure is good and varied.
In all three assessments, one or several of the criteria listed may be present in the student's text. Table 4 gives an overview of the assessments offered by the experienced middle school teacher in relation to the degree to which the students included elaborated NPs. For example, in Essay 1, five of the 30 students received ESN, 15 KA and 10 KMTA. Here all five students who received the lowest assessment (ESN) also produced fewer elaborated NPs than the average. Six of those who received KA produced more elaborated NPs than the average, while nine produced below the average. Finally, of the ten who received the best assessment (KMTA), seven produced above the average and only three below the average. Table 4 indicates that there is a link between assessment/perception and the degree of inclusion of elaborated NPs in narrative writing, even in the stories of very young L2 learners. The impact is most obvious for the ESN assessment (extra support needed). In none of the three essays did the learners whose number of elaborated NPs exceeded the average receive this assessment. It is also quite obvious for KMTA (knowledge more than acceptable), which is an assessment that comparatively few of the learners whose number of elaborated NPs was below average were given (only three of ten in Essay 1; three of 14 in Essay 2; three of eleven in Essay 3). This is in line with Svensson (2018) who arrived at the conclusion that the elaborated NP is the single most important structure to promote narrative writing (see Subsection 2.2.). The fact that there were some students who were given KMTAs (knowledge more than acceptable) despite the number of elaborated NPs in their texts being below average shows, however, that the pre-/post-modified NP is of course not the sole element required to produce more advanced narrative writing. The following tables (5-9) offer the distribution of non-elaborated NPs, elaborated NPs as well as incorrectly and/or unnaturally used NPs. It is clear that already at this early age, L2 learners produce a great many elaborated nominal phrases (correct and natural), and that NP treatment enhances this capacity. In fact, in Essays 1 and 2 (directly after treatment), the number of pre-and/or post-modified noun phrases, reaching a peak in the second story, exceeds the number of noun phrases without any modification. However, it is also clear that the distribution displayed directly after focused instruction-modified heads being in majority (Texts 1 and 2)-cannot be maintained without such scaffolding, Essay 3 instead showing the reverse pattern where non-modified heads are more frequent than ones that are modified. It is therefore also interesting to note that the pattern displayed in Text 3 is contradictory to what appears to come natural in a learner's first language, where, even without treatment, elaborated NPs appear in majority in a story of this descriptive nature (see Table 8 below).     Tables 5 and 6 (second language) also show, in contrast to Table 8 (first language), that the learners produced quite a few modified NPs that were either incorrect and/or unnatural in character. There are two main types. In one case, there was an overuse of epithets, most typically those signalling colour, the exaggerated number also causing the learners to break the rules for the natural order of such elements, as exemplified in:

Results and Discussion
The litel smal brown black hapy dog sitting on the Blo gren black wite red gray purple yello orange flor in a big house.
This type was of course induced by the focused teaching, and tapered off rather quickly, where, in the third essay, it is almost non-existent. The major reason for the extinction of this adjectival overuse is most likely the fact that its unnaturalness was mentioned to the students in connection with Treatment 2, i.e. just before they wrote their second story. Thus, in contrast to Text 1 where adjectives are in majority, Text 2 and especially Text 3 mimic the L1 distribution of nouns, verbs, adjectives and adverbs, nouns displaying the greatest number of tokens/types and adverbs the lowest number of tokens/types. In the other case of incorrect/unnatural use, the learners displayed a tendency to reiterate certain elaborated noun phrases, very often the title offered by the teacher, leaving only one slot in a sentence to be filled, as exemplified in: My best friend is a boy. My best friend is short My best friend is ugly My best friend is tall My best friend is sweet My best friend is nice.
This repetitive use may in fact be the precursor of prefabricated patterns, observed in all types of text (Erman & Warren, 2000;Granger, 1998) irrespective of developmental level (Karlsson, 2002), and can in this case be interpreted as a safety net for students who have not yet acquired the structural level needed. Hence, while such texts may be quite tedious to read (and assessed as weak by teachers), the reiteration of certain nominal groups is most likely not a waste of time from a developmental point of view, as such repetitive writing may be the embryo to what is referred to as syntactic priming whereby the repetition of a structure will pave the way for similar as well as more advanced structures to be acquired (Clealand & Pickering, 2005;Loebell & Bock, 2003) (see Subsection 2.1). With a few learners, this error type remained in all three texts. In the vast majority of cases, however, it too decreased dramatically in number.
The following tables (10-14) offer detailed information regarding the different types of modification in the elaborated NPs. Table 10. Details of pre-and post-modification for Essay 1 after Treatment 1.

PRE-AND POST-MODIFICATION 0%
In addition to an increase in text length (Text 1: 1, 353 words-Text 2: 1, 758 words-Text 3: 2, 412 words, see Table 3), which in itself is an indicator of enhanced narrative writing (Bamberg & Reilly, 1996;Nordenfors, 2011, Svensson, 2018, the most noticeable change can be seen in the distribution of pre-and post-modifying elements. While premodifiers are in majority in all three texts (most typically epithets), as is also the case in the texts written by the students in the two control groups, there is an increase in the number of post-modifiers between the first/second story and the third story. The fact that the unnaturalness of the students' overuse of epithets was mentioned before they wrote their second story may have been what finally forced them to explore other ways of describing the heads of noun phrases, namely by incorporating different types of post-modifiers. It is the use of prepositional phrases that grows the most. Whereas only one PP is used as the single post-modifying element in Text 1 (see Table 10), there are 13 such elements (of the 27 post-modifiers) in Text 3 (see Table 12), implemented by several different learners.
The results also show an increase in the number of relative clauses, from zero as the single post-modifying element in Text 1 (Table 10) to five in Text 3 (Table 12, and from two in combination with pre-modifying elements in Text 1 to four in Text 3, i.e. all in all from two to nine, all restrictive. As discussed in Subsection 2.2, a growth in the number of relative clauses is a clear indicator of growth in (narrative) complexity, i.e. such a change displays a better understanding of how to package information in a more refined way. This increase is an especially interesting observation since, from a Swedish perspective, English relative pronouns may not be categorized as highly accessible items (see Subsection 2.3). Moreover, while the fact that all but one of the relative clauses were zero relatives most likely has to do with transfer from the Swedish language (see Subsection 2.3), it may also be an indicator of learners dreading to have to make a choice among items in a system which is a great deal more complex than their own. In contrast, the fact that Svensson's informant included relative clauses before PPs displays their highly accessible nature in Swedish as an L1 as compared to English as an L2 from a Swedish perspective.
Furthermore, while the two relative clauses observed in Text 1 were produced by the same student, both displaying the same general discourse function (SIT-NEW (situating new referent)), the nine relative clauses in Text 3 were produced by nine different students. Also, whereas eight of these signal general discourse functions, the majority of which again is SIT-NEW, one use signals a narrative discourse function (PRES (presenting main characters)), thus displaying a first step toward being able to construct more complex stories by letting old forms take on new narrative functions. (It is interesting to note that the most common general discourse function in Svensson (2018) is SIT-NEW too, confirming the typically linear development of a story with very young learners in general, be they students of a first or a second language.) The L2 students' use of post-modifying relative clauses may also be compared to what was achieved by the students composing in their L1. Here six of the nine informants, writing two and a half months earlier than the experimental group, already produced eleven such structures in total, although here too the majority displaying general discourse functions.
Furthermore, as is shown in Table 15, Texts 2 and 3, as compared to Text 1, also displays a dramatic growth in coordinated NPs as well as VPs, yet another sign of more advanced narrative writing (Svensson, 2018). Additionally, the first post-modifying when-clause appears in Text 3 (It was a time when it was Christmas). As pointed out in Subsection 2.1, temporal packaging is usually the first type to be made part of texts produced by young learners (Berman & Slobin, 1994). This was also observed in Svensson (2018), indicating L1-L2 similarity in the general developmental trajectory.
Finally, in Text 3 the first formulaic expression -one sopost a time ther was a boy to liked vidio games was used. As mentioned before, it is around the age of nine that such expressions start to appear in the narratives of L1 learners (Svensson, 2018). Further research will have to explore whether a nudge in the form of explicit L2 teaching of NP elaboration is in any way linked to the development of such linguistic forms in a second language.
The reader is here reminded that the treatment the students received took place during a four-week period. While it is certainly possible that some of the development presented above may be attributed to the learners' general cognitive development, it is unlikely that all of it is, especially in the light of the short amount of time between the first and the last test opportunity. It is therefore concluded that it is probable that 'the gentle nudge' in the form of focused NP elaboration accelerated the development of the L2 learners' narrative writing skills. That is, it seems that what has been observed above captures the very first few steps to more enhanced narrative writing in a second language. However, to ascertain statistical confirmation of such a claim, it seems necessary to test much larger student groups.
As presented in Table 16, group level improvement does not, however, imply that all of the students benefitted (equally) from the explicit instruction offered. According to the teacher assessments, if a comparison is made between the first and the last test opportunity, eight students improved in their story writing skills. It may be that one of the next logical steps in these learners' overall trajectory, cognitively as well as linguistically, in accordance with the Teachability Hypothesis as well as the Zone of Proximal Development (see Subsection 2.1), was to pay attention to NP modification, thus making these students especially attuned to the focus of the instruction at hand. 16 students received the same assessment on the first and the last test opportunity. As for those receiving KMTA (knowledge more than acceptable), these learners appear, at least on the scale implemented in the present study, to have maxed out. That is, ceiling effects may be at play, the highest assessment not being able to capture the development of the most advanced learners. The opposite can be said for those students receiving the poorest assessment on both occasions. That is, there may also be floor effects at play, ESN (extra support needed) not being able to capture improvement among the weakest learners. Also, as the poorest assessment in itself incorporates a great span of different students (see Subsection 3.4), it may be that these particular learners were still stuck in a mindset geared towards spoken language, dealing with precursors to elaborated NPs (see Subsection 2.2) (Plunkett & Strömqvist, 1992), making NP elaboration too much to handle for them at that point in time. Lastly, the group consisting of students who received KA (knowledge acceptable) on both test opportunities may have been in a state of turbulence in which they were trying to sort out a number of different linguistic problems, and presenting them with NP elaboration was just one aspect too many. The fact that the KA assessment in particular involves stories that display number of NPs above as well as below average (see Table 16) points in this direction.
A turbulent state may also be the reason why the focused NP elaboration appears to have been detrimental to six of the students, all (superficially?) regressing between the first and the last test opportunity. However, backsliding (Selinker, 1972) is a common feature of all parts of second language acquisition, as sorting out of various aspects may be going on simultaneously, and should therefore not necessarily always be the cause of too much concern. In fact, it may be that these students may later (subconsciously) return to what they were taught with more positive results.
The findings indicate that, if L2 students who are in a developmental stage where they are susceptible to pre-and post-modification of nominal phrases could be singled out, focused NP elaboration could set things in motion and develop these learners' narrative writing skills in an enhanced manner. To determine what that level of susceptibility incorporates, more research in this area is definitely needed.

Summary and Limitations to the Study
In the present investigation, 30 Swedish middle school students (aged 10-11) were, with intervals of two weeks, asked to write three narrative texts in L2 English. With the aim of enhancing their narrative writing skills, the learners received focused teaching of pre-and post-modification of nominal phrases prior to the first two compositions, while no such instruction was offered before the third text. These students were also compared with two control groups, one writing in Swedish (L1) and one writing in English (L2), neither being offered treatment before doing so.
In an initial stage, the students displayed an overuse of adjectives in pre-modifying position, assumed to have been induced by the instruction offered, as well as repetitive patterns of identical elaborated noun phrases. The mentioning of the unnaturalness of the overuse of pre-modifying adjectives before the second story was composed seems to have ignited attempts to instead elaborate by using post-modification. Here the use of prefabricated patterns probably also paved the way, as syntactic priming is a way for learners to slowly acquire more complex structures. As a result, the students' stories did not only increase in text length but also displayed a growth in post-modifying prepositional phrases and relative clauses. Moreover, while the majority of these relative clauses displayed general discourse functions, there were also signs of the learners starting to use relative clauses having narrative discourse functions, a more complex type, and designed not only to form cohesive and coherent texts but imaginative narratives. Text 3, in contrast to Text 1, also included more coordinated NPs and VPs, as well as the first when-clause and the first example of a formulaic sequence, all signs of more advanced text construction. However, as indicated by the teacher assessments, not all of the L2 learners appeared (equally) susceptible to the instruction of NP elaboration, indicating that more research is needed to be able to single out those students that are in a stage where they can benefit from such teaching.
As this study is the first to deal with the effects of focused NP elaboration on young L2 learners' narrative writing skills, it inexorably suffers from some limitations. For example, had larger groups of students of different ages been asked to write several essays during a longer period of time, the investigation would not only have been able to capture the difference in use between pre-and post-modifying elements, but might also have been able to more clearly offer a picture of the use and development of the individual elements among pre-modifiers (deictics, epithets, and classifiers) on the one hand and post-modifiers (PPs, relative clauses and other types of clauses) on the other hand. The study might also have been able to reveal reciprocal links between the various pre-and postmodifying elements.
It would of course also have been better if the student group who received L2 instruction could have written a text prior to treatment. Involving a smaller control group for this purpose may have affected the data in a number of ways. Here the instructions themselves should also be mentioned, where, for instance, other topics for the students' narrative assignments might have yielded different results. Larger test groups in future studies may help resolve the effects of such factors.