The main focus of this survey is emotion-dependent variation of automatically generated language, or Emotional NLG for short. Although no longer young, this is still an emerging field of research that concerns itself with varying language output (in NLG systems) to reflect different emotions. However, the survey also includes a literature review of the closely related (and similarly small) field of Verbal Emotion Identification which develops methods for spotting emotional signifiers in language. The reason for the inclusion is that there is substantial overlap between the two fields: (i) both seek to identify lexical, syntactic and semantic properties that are correlated with specific emotions; (ii) both have employed psychologically inspired models of emotion. Furthermore, as will be argued, emotion identification methodology can provide quantitative, empirical grounding for emotional NLG.