The well-known modern dimensional models of affect include two or three dimensions. They are typically based on self-reports using English emotion terms. It remains unclear to what extent these models can be applied to different cultures and languages. The present study is aimed at finding the dimensions underlying the descriptions of emotional states in Russian language and suggests a structural model of affect based on these findings. At the first stage, a comprehensive list of Russian nouns denoting emotional states was compiled. It comprised 330 words and was reduced to a list of 56 words. At the second stage, participants rated their emotional states using this list. The exploratory factor analysis yielded three dimensions that underlay participants’ descriptions of the emotional states – negative affect with low activation, positive affect with high activation, and tension. This structure has at least three notable features. First, valence is not a bipolar factor like in some other structural models of affect. Instead, it splits into two orthogonal factors. Second, valence is somewhat related to arousal, namely positive affect is associated with high arousal and negative affect with low arousal. Third, emotional states related to tension and uncertainty form a separate factor. This factor presumably reflects pure tension independent from any valence. This model can be used for developing Russian-based measures for assessment of mood.