Learning a new language to meet one’s everyday needs, for instance in professional or educational contexts, is doubtlessly a challenging task. How this learning process is experienced and managed, and what can hinder or facilitate it, are questions that researchers and practitioners are concerned with. These overarching issues are addressed in this dissertation: How do second language learners of Norwegian demonstrate and develop linguistic complexity? To investigate all of the features of linguistic complexity is not possible within the scope of a dissertation, so the overarching issue has been narrowed down to the following research question: How can complexity and complexity development of the Noun Phrase as a construction be analysed and described in learner language? I have chosen a construction grammar approach to this investigation (Bybee, 2008, 2013; Goldberg, 1995, 2003, 2013; Goldberg & Casenhiser, 2008; Hoffmann, 2013), and as a result of this choice I refer to the Noun Phrase as a construction: a Noun Phrase Construction. In order to conduct this investigation, a corpus of texts was collected consisting of answers to the task ‘written production’ at the Test of Norwegian (no.: Norskprøven) for adult second language learners, from the summer of 2016. The corpus contains 60 texts (N1-N60), distributed equally across the four CEFR reference levels A1, A2, B1 and B2. Additionally, five texts written by L1 users of Norwegian were collected and used as a control corpus (N61-N65). The corpus contains 23,116 words. A total of 4,801 occurrences of the Noun Phrase Construction were identified (NP-construction): 2,269 consisted of a single noun, and 2,532 consisted of a noun and one or more modifiers. The corpus, with tags, counts and classifications, as well as the significance tests, are available by request: firstname.lastname@example.org or email@example.com. Ten considerations that were made in connection with marking the corpus are summarized in the dissertation’s methodical chapter (chapter 3) and can be found in their entirety in appendix 4. The results of the significance tests are presented in appendix 5, and 20 systematized overviews of selected occurrences can be found in appendix 6. In order to study learner development, one needs a measurement tool that accounts for stages of proficiency. In the dissertation, the reference levels A1, A2, B1 and B2, according to how they are described in A Common European Framework of Reference for Languages (Council of Europe, 2001), hereafter CEFR, served as the independent variables for the comparison. The research design can be regarded as a corpus linguistic study, even though the corpus size is relatively modest. I have investigated how NP-constructions are used at each reference level (chapter 6), and I have identified tendencies and development patterns across the four reference levels (chapters 7 and 8). At a superordinate level, the NP-Construction can be modelled as such: [(_) + N + (_)]. In addition to this abstraction level, two further levels of analysis or development can be identified. I have chosen to call them the systematization level and the experience level. Following the dissertation’s analyses, results, and findings, I have proposed to connect these three levels in what I have called a construction grammar learning pyramid. The analyses, results and findings show that learners activate linguistic resources in a creative way as they navigate, consciously or unconsciously, between the experience level, the systematization level and the abstraction level in their learning process. The construction grammar learning pyramid shows how constructions at the experience level gradually establish foundations for constructions at the systematization level that eventually can be abstracted in linguistic structures at the abstraction level. The pyramid illustrates how, within a construction grammar approach, exemplars can create systems that then shape frames that new constructions or exemplars of constructions can be created within. The analyses in the dissertation contribute to demonstrating that the development represented by the pyramid is neither one-dimensional nor linear, but rather is constituted by many parallel developments that build on one another, and that, in some cases, appear to predetermine or assume one other. When one developmental trait seems to predetermine another trait, the development is called emergentistic (Ellis, 2006). Through my analyses, I have found examples of such traits, and I will present them in chapters 7 and 8. In my analyses of complexity and complexity development, the systematization level has been especially important. The NP- construction at this level can be presented as such: [(definer) + (modifier)+ N + (definer) + (modifier)]. This construction offers 15 possible combinations: [definer + N] as in [en gutt] and [tre gutter], [describer + N] as in [God helg!] and [fine hus] etc. An important finding in the dissertation is that the development of which possible combinations are utilised is not arbitrary, but rather seems to follow a systematic, emergentistic complexity development. The development seems to persist until the language learners reach the B2-level. The eight combinations that are used correspond with the combinations used in the control corpus. The eight combinations are: [definer + N]: litt vann [describer + N]: god lønn [N + definer]: drømmen min [N + describer]: pølse med brød [definer + describer + N]: en uheldig familie [definer + N + describer]: et liv som er verdt å leve [describer + N + describer]: positivt svar fra dere [definer + describer + N + describer]: en liten gutt som leker med en katt In order to operationalize the investigation of complexity and complexity development at the systematization level, I have developed a series of measurements for complexity analyses. These measurements have commonalities with previous research on complexity in learner language (Bulté & Housen, 2012). I have chosen to sort them in the following main groups: length complexity, morphological complexity, syntactic complexity, content complexity and referential complexity. The complexity measurements make possible an exploratory approach to the data analysis, and they are therefore well suited to identifying how complexity manifests both within and across the reference levels A1, A2, B1 and B2. At the same time, I have investigated the relation between complexity and accuracy, in line with research within the CAF-tradition (complexity, accuracy and fluency) (Alanen et al., 2010; Bulté & Housen, 2012; Housen & Kuiken, 2009; Lambert & Kormos, 2014). Research within the CAF-tradition has pointed out that language proficiency in the three dimensions do not necessarily develop in tandem: it can be a procedural challenge for a learner to work, consciously or unconsciously, with the complexity dimension and the accuracy dimension at the same time. My analyses support this claim. NP-constructions at the experience level are also examined in the dissertation. The reason for naming the most basic level in the learning pyramid the experience level is that constructions at this level seem to be conceived as (more or less) singular entities that the learner, consciously or unconsciously, gradually systemizes. The transition between the experience level and the systematization level can therefore be perceived (for the learner as well for the teacher) as ever-changing and unstructured. For the purpose of investigating whether this transition really is «ever-changing and unstructured», it is essential not only to focus on the systematization level, but also the experience level. An investigation with such a focus can provide valuable insight into how the transition between experience and systematization manifests, and what characterizes it. I investigate the development of six constructions at the experience level in chapter 8 of the dissertation. These constructions can be presented as such: Two part-quantitative constructions: [number + X-er] and [part- quantifier + X-er] Two possessive constructions: [possessive + X] og [X-en/-a/-et/-ene+ possessive] Double determiner definitt construction: [den/det/de + adjective + X- en/-a/-et/-ene] NP-Constructions with a that-clause as a postmodifying determiner: [X som (gjør) Y] Since the NP-construction is the main topic of the dissertation, all of the six investigated constructions are NP-constructions, which means that X in these constructions will always consist of a noun. Even though the constructions are located at the experience level of the construction grammar learning pyramid, they hold the potential for complexity. For instance, a premodifier (an adjective) can be included in a possessive construction: [min gode venn], or a postmodifying prepositional phrase can be included in a part-quantitative construction: [mange pølser med brød]. How learners use (or do not use) this potential for complexity, is also investigated. Overall, the dissertation’s investigations and analyses have demonstrated that the development of the NP-construction’s complexity is emergen- tistic, creative and systematic. The findings have also indicated how this development takes place. Thus, studies of complexity and complexity development that use construction grammar as their point of departure can offer important insights into the development and nature of language learning and can consequently contribute to developing didactic practices within second language teaching.