Spoken emotion and semantic emotion are two components of emotion expression. In human conversation, emotions expressed by these two modalities are similar in healthy individuals. However, rich evidence documents that depression might affect emotional expression. Nevertheless, the consistency between spoken and semantic emotion in depressed patients has rarely been studied previously. In the present study, we investigated the consistency between emotions expressed by acoustical features and text content in depressed and healthy individuals during natural conversations. It was found that depressed patients tended to talk about negative topics in a neutral emotional tone and talk about neutral or positive topics in a depressed tone. These findings suggest that depression not only affects the emotion expression of a single modality but also results in an inconsistency between emotions expressed by these two modalities.