Following the detection-based approach, this article detects statistically significant frequency differences between the data of written Finnish by learners from various language backgrounds. It analyses crosslinguistic influences in a data-driven manner, as the analysis focuses on the morphological forms and their combinations (n-grams) that prove to be the best predictors of differing first languages. Following the methodology applied – key structure analysis – the article then goes on to analyse the found n-grams in terms of their inner and cotextual variation in order to find out which linguistic phenomenon actually distinguishes the subsets of data. The results show several quantitative differences that may be due to the crosslinguistic influences and they were all detected in a data-driven manner without hypotheses of potential differences. The method can be useful especially in finding and analysing elusive crosslinguistic influences that cannot be interpreted to be transferred directly from the respective first languages.

Key words: Finnish as a second language, crosslinguistic influence, detection-based

approach, key word analysis, key structure analysis