To shed light on these issues, in this paper we aim to provide a comprehensive body of evidence regarding the effectiveness and limitations of different sentence features. Our main contribution is of an empirical nature: the focus of this paper is to report an extensive set of experiments planned to evaluate the relative effectiveness of different sentence features for subjectivity classification and polarity classification. Specifically, we designed a general experimental setting – with two SA tasks and different sources of data – to jointly evaluate features that have otherwise been tested independently. To the best of our knowledge, this is the first study of this kind for SA at sentence level.