Abstract We introduce a parallel treebank based on SweLL, the Swedish Learner Language corpus. Annotation follows the Universal Dependencies (UD) standard, whose descriptive nature and focus on cross-linguistic comparability make it well suited for the analysis of learner material. We discuss the annotation challenges arising from the idiosyncrasies of L2 Swedish, present the internal guidelines that were developed to address them, evaluate the annotation quality of the treebank and investigate which factors affect the difficulty of the annotation task. The resource currently consists of 510 gold-annotated sentence-correction pairs and is in the process of being expanded.