Spoken and written language are two modes of language. When learners aim at higher skill levels, the expected outcome of successful second language learning is usually to become a fluent speaker and writer who can produce accurate and complex language in the target language. There is an axiomatic difference between speech and writing, but together they form the essential parts of learners’ L2 skills. The two modes have their own characteristics, and there are differences between native and nonnative language use. For instance, hesitations and pauses are not visible in the end result of the writing process, but they are characteristic of nonnative spoken language use. The present study is based on the analysis of L2 English spoken and written productions of 18 L1 Finnish learners with focus on syntactic complexity. As earlier spoken language segmentation units mostly come from fluency studies, we conducted an experiment with a new unit, the U-unit, and examined how using this unit as the basis of spoken language segmentation affects the results. According to the analysis, written language was more complex than spoken language. However, the difference in the level of complexity was greatest when the traditional units, T-units and AS-units, were used in segmenting the data. Using the U-unit revealed that spoken language may, in fact, be closer to written language in its syntactic complexity than earlier studies had suggested. Therefore, further research is needed to discover whether the differences in spoken and written learner language are primarily due to the nature of these modes or, rather, to the units and measures used in the analysis.

SammanfattningI denna studie undersöks hur syntaktisk komplexitet och korrekthet skiljer sig på två olika färdighetsnivåer i skriftlig inlärarsvenska samt vilken relation som föreligger mellan komplexitet och korrekthet på dessa färdighetsnivåer. Enligt tidigare studier ökar både komplexiteten och korrektheten i syntaktiska strukturer i samband med ökade kunskaper i målspråket, men det finns också studier som har visat på motstridiga resultat (Wolfe-Quintero et al., 1998). Vidare är det oklart om relationen mellan dessa komponenter är konkurrerande (Skehan, 1998) eller stödjande (Robinson, 2001).Materialet i denna studie består av 31 uppsatser skrivna av finska universitetsstudenter. Fokuset ligger på strukturernas yttre komplexitet och följande komplexitetsmått används i analysen: den genomsnittliga längden på T-enheter och satser, andelen bisatser och förekomsten av bisatser med satsadverbial och topikaliserade strukturer. Korrekthetsaspekten avgränsas till fel i ordföljden och mäts med andelen felfria satser, felfria bisatser och felfria bisatser med satsadverbial samt förekomsten av inversion jämfört med det totala antalet topikaliserade strukturer.Resultaten tyder på att färdighetsnivån huvudsakligen påverkar korrekthetsaspekten medan komplexiteten också kan påverkas av andra faktorer, såsom inlärarens allmänna kännedom om strukturer i målspråket. Något överraskande visar resultaten inga statistiskt signifikanta korrelationer mellan komplexitet och korrekthet. Därmed visar resultaten inget stöd för vare sig konkurrerande eller stödjande samband mellan dessa komponenter, med eventuellt undantag för en tendens till konkurrerande relation mellan bisatser med satsadverbial och korrekthet på den högre färdighetsnivån och mellan topikaliserade strukturer och korrekthet på den lägre nivån.

This study examines how proficiency level affects syntactic complexity in written and spoken L2 Swedish. Spoken production can be more challenging for learners at lower proficiency levels due to the limited resources for processing the target language (Skehan 1998), whereas the possibility for more extensive planning during the writing process enables the production of syntactically complex structures in writing (Skehan & Foster 2012, Tavakoli 2014). Automaticity at higher proficiency levels reduces the cognitive load that the spoken production imposes on learners and facilitates the production of complex structures in speech (Leonard & Shea 2017). However, few studies have directly compared complexity in written and spoken L2 production, and earlier studies have resulted in contradictory results (Kuiken & Vedder 2012b:364–365). The data in this study consist of monologues and essays produced by 31 Finnish university students. Some subjects study Scandinavian languages as their major, while others take part in a Swedish course at the Centre for Language and Communication Studies. The analysis is based on surface-level complexity measures, such as the mean length of production units and the dependent clause ratio. The results show that proficiency plays an important role especially in spoken production and, therefore, also in the differences between written and spoken production. At the lower proficiency level, written production is significantly more complex when it comes to the mean length of T- and AS-units, the dependent clause ratio, and the ratio of dependent clauses with a sentence adverbial. Surprisingly, the mean clause length shows higher complexity in speech at the lower proficiency level. At the higher proficiency level, there are no statistically significant differences between the modalities. These results indicate that the cognitively more demanding task of producing L2 speech can lead to the use of simple structures at the lower proficiency level, resulting in a greater difference between written and spoken complexity. The learners at the higher proficiency level seem to be able to produce complex structures also in the spoken mode, and therefore, there are no significant differences between written and spoken complexity at this level.