Background: Despite advances in the development and testing of therapies for verb retrieval impairments in aphasia, generalisation effects of treatment remain a challenge. Semantic Feature Analysis (SFA) is a word retrieval treatment that has been reported to result in generalised responding to untrained object names with persons with aphasia. The theorised therapeutic mechanisms of SFA appeared to be appropriate for facilitating retrieval of trained and untrained action names. Aims: This investigation was designed to extend pilot research in which SFA was applied to verb retrieval. The primary purpose of the current study was to examine the acquisition and response generalisation effects of SFA applied to action naming with four persons with chronic aphasia. Additional purposes were to examine changes in production of content in discourse and to explore the correspondence of accuracy of naming during treatment to probe performance. Methods & Procedures: SFA was modified slightly to be appropriate for application to action naming as opposed to object naming; several feature categories were changed, but all other procedures were retained. Treatment was applied sequentially to two sets of action names in the context of multiple baseline designs across behaviours and participants. Accuracy of naming of trained and untrained actions in probes was measured repeatedly throughout all phases of the design. Production of correct information units (CIUs) in discourse was measured prior to and following treatment. The relationship of probe-naming performance to naming performance during treatment sessions was examined using correlational analyses. Outcomes & Results: Increased accuracy of naming of trained action names was associated with treatment for three of the four participants. The remaining participant did not demonstrate improvement in naming on probes, despite some gains during treatment. Generalisation to untrained action names did not occur for any of the participants. Increases in CIU production were observed for only one of the participants. For the participants with positive naming outcomes, probe performance correlated well with naming performance during treatment. For the participant who demonstrated some improvements in treatment, but did not show gains in naming on probes, weak correlations were obtained. Conclusions: SFA appears to have potential for promoting improved action naming in aphasia. However, more research is warranted to explore treatment modifications to promote generalisation. Correlational analyses indicated that gains in naming during treatment may not always be reflected in probe performance and thus, require verification through probing in non-treatment conditions.