In everyday life, people use language to communicate evaluative messages about social categories. A selection bias in language across two social dimensions not previously integrated was examined: a self-inclusive/self-exclusive dimension and an individual/collective dimension. Pronouns as markers for social categories were adopted (I, We, He/She, and They), and a new measure was developed (the Evaluative Sentence Generating task) to investigate the evaluative context selected for the pronouns. Results demonstrate that individuals select a more positive context for self-inclusive than self-exclusive pronouns and a more positive contexts for individual than collective pronouns. However, in an interpersonal context, evaluative differences between I and We diminished, whereas in an intergroup condition the evaluative gap between self-inclusive and self-exclusive pronouns was magnified.