In this paper we describe the system used to participate in the sub task 5b in the Phrasal Semantics challenge (task 5) in SemEval 2013. This sub task consists in discriminating literal and figurative usage of phrases with compositional and non-compositional meanings in context. The proposed approach is based on part-of-speech tags, stylistic features and distributional statistics gathered from the same development-training-test text collection. The system obtained a relative improvement in accuracy against the most-frequentclass baseline of 49.8% in the “unseen contexts” (LexSample) setting and 8.5% in “unseen phrases” (AllWords).