This paper presents some preliminary results of the SIDE project, which aims at investigatingthe emotional content of idioms from both a behavioral and computational point of view. In this first work, we collected affective ratings for a set of 45 Italian verb-noun idioms and 45 Italian non-idiomatic verb-noun pairs and then performed a computational study to analyze the role of linguistic information in determining the affective rating of idioms and non-idioms. To this end, we exploited an existing vector-based computational model to calculate a distributional polarity score. The correlation analysis between such a score and human-rated valence revealed a high correlation among the variables as well as a significantly lower polarity score of idioms with respect to non-idioms reflecting their stronger negative content.