This chapter compares frequent four-word lexical bundles in a learner corpus (VESPA) and a native speaker corpus (BAWE), both representing novice academic writing. The frequencies and dispersion of bundles in the two corpora reveal patterns of both over- and underuse among the learners. The learners are shown to use some bundles very frequently, but frequencies drop more sharply than in the native corpus. The dispersion of the frequent bundles tends to be broader in the native speaker corpus. In a closer scrutiny of four selected bundles the novice-expert dimension is addressed by consulting a corpus of published research articles. Contrasts between English and Norwegian are also considered in order to explain the learners’ apparently non-native usage. Some of the most overused bundles seem to have been generalized by the learners to fit into contexts where native speakers rarely use them; these can be described as ‘phraseological teddy bears’. Pedagogical applications of the results should start from the underused items in order to broaden the phraseological repertoire of the learners.