Background: Semantic feature analysis (SFA) is a prominent treatment for the word retrieval deficits of aphasia. Generalisation of improved word retrieval on picture naming to discourse production has been an important factor for evaluating the effectiveness of SFA treatment. Unfortunately, generalisation of word retrieval improvements to discourse production following SFA has been modest. Aims: Because of the previous, albeit limited, success of SFA in producing improved word retrieval for discourse we further examined the utility of SFA for reducing noun and verb retrieval failures in aphasic discourse. Rather than use SFA as a means for improving generalisation of picture naming or as a compensatory strategy for lexical failures during discourse, we applied SFA as an a priori means to reduce the frequency of word retrieval failures in discourse. Methods and Procedures: Semantic feature analysis was applied to object and action word retrieval failures appearing during picture descriptions and procedural questions by two participants with anomic aphasia. A single case time-series design across behaviours with replication was used to assess changes in discourse production as well as generalisation of treatment effects to untrained pictures resulting from SFA. Outcomes and Results: Increases were observed in verbal productivity for both participants, while the informativeness of the participant's discourse, as measured by correct information unit analyses, also improved. Minimal changes were observed in the frequency and type of word-finding behaviours evinced by the participants; this finding was attributed to a masking effect arising from the participants' increased quantity of verbal output. Evidence was also found that targeting word finding behaviours in connected speech generalised to naming of untrained object and action pictures. Conclusions: The changes effected by this discourse-based approach to SFA were as robust and as consistent as has been achieved previously with SFA treatment. The choice to use a discourse-based versus a picture-based approach to SFA treatment might be based on the ecological validity of the discourse-based approach.